Eysenck Personality Questionnaire 15

Structural and Functional Equivalence of the Eysenck

Personality Questionnaire Within and Between Countries 

Dianne A. van Hemert

Fons J. R. van de Vijver

Ype H. Poortinga

Tilburg University, The Netherlands

James Georgas

University of Athens, Greece


 Eysenck Personality Questionnaire 16

Abstract
The question was examined as to whether scores at the individual level and scores
at the country level on the four scales of the Eysenck Personality Questionnaire
(EPQ) have the same psychological meaning. Using data of 24 countries, it was
found that the EPQ has different factorial structures at both levels. Both the Lie scale
and the Psychoticism scale were shown to jeopardize cross-level equivalence. For
further exploration of the meaning of the EPQ scales within countries and between
countries country-level correlations were calculated with a variety of country
characteristics such as Gross National Product, political indices, religiosity,
Hofstede's measures, and subjective well-being. Significant findings for 38 countries
included correlations of the EPQ scales with Hofstede’s Masculinity, Diener’s
Subjective Well-Being, religiosity, the number of deaths in a country due to political
violence, and bribery. The most striking finding was a substantial negative correlation
of the Lie scale with Gross National Product and other wealth-related indices.
 Eysenck Personality Questionnaire 17

Structure and Score Levels of the Eysenck Personality Questionnaire across

Individuals and Countries
The Eysenck Personality Questionnaire was published in 1975 (Eysenck &
Eysenck, 1975) and consists of four scales. The Psychoticism scale (EPQ-P) was
designed to measure tough-mindedness, the Extraversion scale (EPQ-E) was meant
to measure extraversion versus introversion, and the Neuroticism scale (EPQ-N) was
constructed to measure emotionality or emotional instability. The interpretation of the
Lie scale (EPQ-L) is less straightforward. From the beginning, Eysenck and Eysenck
(1976) have acknowledged that the EPQ-L, besides (or as part of) measuring a
tendency to fake good, reflects a stable personality characteristic, namely social
conformity. This was confirmed by McCrae and Costa (1983), who found the EPQ-L
to be substantially related to neuroticism and extraversion. They concluded that
social desirability scales more likely measure a personality characteristic than a
response set (see also Ones, Viswesvaran & Reiss, 1996, for a similar conclusion).
Šipka (1988) questioned the use of the EPQ-L as a dimension of personality in cross-
cultural comparison until there is more clarity as to its nature.
Over the years, many cross-cultural studies have been done with the EPQ. In
studying the construct equivalence of the EPQ scales across countries two general
strategies can be followed. First, one can research the internal structure of a concept
(structural equivalence). Barrett, Petrides, Eysenck, and Eysenck (1998) established
structural equivalence of the EPQ over 34 countries, which means that the factorial
structure is (nearly) identical in each of these countries. This is an important finding.
At the same time, the question remains unanswered whether differences in scores
between countries have the same psychological meaning as differences in scores
within countries. Do individual differences in, say, extraversion have the same
meaning as country differences or do the latter differences have to be accounted for
by other factors, such as cross-culturally different response sets? The second way to
assess cross-cultural construct equivalence is to examine (parts of) the nomological
networks of the instrument in the different cultures (Van de Vijver & Leung, 1997).
This strategy can be said to focus on functional equivalence, as it refers to the
functional context of the concept.
In the present study two issues are addressed. The first concerns the
structural equivalence of the Eysenck Personality Questionnaire at within- and
between-country levels. The second issue attends to the functional equivalence
through examining at both levels relationships with variables that can be seen as part
of the nomological network of the EPQ scales. Secondary analyses of previously
published studies are used to address these issues.
 Eysenck Personality Questionnaire 18

Structural Equivalence of the EPQ

To evaluate whether the EPQ scales provide useful dimensions to describe
individual differences across countries, the structure found in various countries can
be compared to a reference population using target rotation (Van de Vijver & Leung,
1997). Quite a few studies have done this, using an English sample as reference
(e.g., Barrett & Eysenck, 1984; Eysenck, Barrett, Spielberger, Evans, & Eysenck,
1986; Eysenck & Haraldsson, 1983). In these studies the EPQ was administered in
some country and the factorial structure was compared to the UK norm groups, using
a procedure described by Kaiser, Hunka, and Bianchini (1971) for factor comparison
(e.g., Eysenck & Eysenck, 1983). It was found that this comparison method could
yield misleadingly high coefficients of factorial similarity (Bijnen, Van der Net, &
Poortinga, 1986; Bijnen & Poortinga, 1988). Moreover, simulation studies have
shown that target rotation procedures such as those cited by Eysenck and Eysenck
(1983) do not have a high power to detect biased items (Van de Vijver & Poortinga,
1994). More definite evidence was derived by Barrett et al. (1998) using an improved
comparison procedure, which also showed factorial similarity of the EPQ across 34
countries.
There is a second aspect to structural equivalence in cross-cultural
comparisons, namely whether it holds for both individuals and populations. A
necessary condition for equality of psychological meaning of individual (within-
country) and country (between-country) scores is the equality of the factor structure
at these two levels.1 Dissimilarity of within-country and between-country factor
structures means that different underlying factors account for scores at the two
levels. An illustrative example of different relations at different levels was described
by Myers and Diener (1996). In most nations there is hardly a relationship between
income and happiness. Yet, people in rich countries are generally happier than
people in less prosperous countries. A way to establish equivalence of constructs at
different levels was proposed by Muthén (1991, 1994). He describes a multilevel
factor analysis or covariance structure analysis, based on a comparison of the pooled
within-sample structure with the between samples structure. The pooled within-
sample covariance matrix is an average of the covariance matrices of the separate
samples, each weighted according to sample size. The between-samples matrix is
computed on the basis of the aggregated sample means of the various variables.
Muthén’s procedure, developed for confirmatory factor models, can be easily
extended to exploratory factor analytic models (Van de Vijver & Poortinga, 2001).
The strategy amounts to first factor analyzing the pooled within-country correlation
matrix and the between-country correlation matrix and then evaluating the agreement
 Eysenck Personality Questionnaire 19

of the loadings after rotation of the two solutions. The factorial agreement can be
evaluated by a congruence coefficient. A second method to assess factor similarity
involves a bootstrap procedure that makes it possible to examine the statistical
significance of congruence coefficients (Chan, Ho, Leung, Chan, & Yung, 1999).
Both procedures are discussed in more detail in the Method section.
Functional Equivalence of the EPQ
An evaluation of the nomological network of the EPQ requires correlations
between context variables and the EPQ scales at individual and country level. Many
studies have investigated EPQ correlates within countries and some studies have
done the same across countries. However, no studies have attempted to compare
correlates of the EPQ scales at both levels. Because of the amount of within-country
studies, the present literature review is limited to studies concerning correlates
across countries. For a summary of findings at the individual level the reader is
referred to Table 4.
Lynn has done extensive research on the relationship between the EPQ
scales, Hofstede’s measures, and demographic variables at country level (e.g., Lynn,
1971, 1981; Lynn & Hampson, 1975, 1977). Lynn (1981) postulated that cross-
cultural differences in neuroticism are related to differences in stress. He argues that
stress is linked to political, social, and economic instability, as well as to war and
climate. Also, Lynn indicates that higher levels of extraversion were found in more
affluent nations. Lynn (1971) calculated rank correlations for 11 countries between
anxiety, as measured by Cattell’s 16 Personality Factor Test, and various country
variables. Anxiety can be regarded as related to neuroticism. Significant positive
correlations were found with alcoholism and average temperature in the hottest
month of the year, and a significant negative correlation with rates of psychosis. Lynn
(1981) reported a highly significant negative correlation (r = -.62) between EPQ-P
and per capita income across 12 nations.
In a recent study, Lynn and Martin (1995) reported national means of the E,
N, and P scales for 37 countries. These means were correlated with demographic
variables such as national rates of suicide, homicide and alcoholism, with economic
data such as per capita income, and with work attitude data such as work ethic,
competitiveness and an anxiety index. The latter was derived from Hofstede (1976),
who measured anxiety with a single item, namely “How often do you feel nervous or
tense at work?” Significant negative correlations were found between psychoticism
and work ethic, and between extraversion and rate of suicide, while positive
correlations were reported between neuroticism and Hofstede’s anxiety index, and
 Eysenck Personality Questionnaire 20

between extraversion and rate of homicide. No significant correlations were found

between the EPQ scales and per capita income.
Lester (1988) found in a sample of 18 industrialized nations anxiety scores to
be significantly correlated with suicide rate (r = .45), and extraversion scores with
homicide rates (r = .60). His study of neuroticism and extraversion in 32 nations
(Lester, 2000) rendered similar results as Lynn's (1971; Lynn & Hampson, 1975)
studies.
Arrindell et al. (1997) correlated subjective well-being with various
countrylevel variables, among which EPQ-P, EPQ-E and EPQ-N (obtained from Lynn
& Martin, 1995), across 20 countries. They found subjective well-being to be
negatively correlated with neuroticism and psychoticism. Positive correlations were
found between neuroticism and Hofstede’s (1980) dimensions of Power Distance,
Masculinity and Uncertainty Avoidance, as well as for psychoticism and Power
Distance.
Ones et al. (1996) meta-analyzed data on social desirability and correlated
the data with (among other variables) the Big Five personality dimensions. Social
desirability was found to be related to emotional stability (r = .37), conscientiousness
(r = .20), and years of education (r = -.18). In a survey Ross and Mirowsky (1984)
found that respondents in the USA and Mexico with lower socio-economic positions,
higher age, or Mexican descent scored higher in acquiescence and social
desirability. They concluded that less powerful social groups give more socially
desirable responses. They state that the same tendency occurs among social groups
that stress the importance of keeping up a proper image, because socially desirable
responses may be seen as strategies for presenting a good face.
Warnecke et al. (1997) reported higher social desirability scores in the USA
among both African American and Mexican American respondents compared to non-
Hispanic Whites, after controlling for gender, age, education, and income. Other
studies also have shown higher social desirability scores among Blacks than among
Whites in the USA (e.g. Crandall, Crandall, & Katkovsky, 1965; Edwards & Riordan,
1994; Fisher, 1967; Johnson & Van de Vijver, 2000; Klassen, Hornstra, & Anderson,
1975).
Williams, Satterwhite, and Saiz (1998) measured favorability (which appears
to be closely related to social desirability) in ten countries with the Adjective Check
List (Gough & Heilbrun, 1965). This 300-item list of person-descriptive terms was
presented to samples of students, who were asked to indicate the favorability of each
adjective on a 5-point scale. Reanalyzing these data, Poortinga and Van de Vijver
 Eysenck Personality Questionnaire 21

(2000; see also Johnson & Van de Vijver, 2000) found a significant correlation (r =
-.84) between the average favorability of all items and a socioeconomic index.
Bond and Smith (1996) performed a meta-analysis of studies using Asch’s
line judgment task, which can be considered an alternative measure for social
conformity. Across 17 countries they found significantly higher levels of conformity for
countries with higher collectivism scores, indicated by three surveys assessing a
country’s individualism or collectivism, among which Hofstede's (1980) index.
A number of studies have attempted to find a relationship between the EPQ
scales and religiosity or attitudes towards religion. Unfortunately, correlations are
generally calculated only within countries. Although the present section focuses on
correlations across countries, some of these within-country studies are mentioned in
order to formulate expectations concerning the relationship between religiosity and
personality across countries. In most studies a negative relation between
psychoticism and religiosity is found (Francis, 1992; Heaven, 1990; Kay, 1981; Lewis
& Maltby, 1995; Svensen, White, & Caird, 1992). Francis, Lewis, Brown, Philipchalk,
and Lester (1995) compared students from the UK, USA, Canada, and Australia and
concluded that psychoticism is fundamental to individual differences in religiosity.
Extraversion has been reported to be negatively correlated to religiosity (Francis,
Pearson, Carter, & Kay; 1981), but the relationship is not consistent (e.g., Caird,
1987). For EPQ-N sometimes positive and sometimes negative correlations are
reported. Finally, Lie scale scores have been found to be positively correlated with
religiosity (Francis, 1985).
In summary, although several studies have addressed the structural
equivalence of the EPQ between two or more countries, they only considered
equivalence within countries. However, one cannot be certain that differences in EPQ
scores at the individual level have the same meaning at the country level. Thus, our
first research question addresses the structural equivalence of the EPQ within and
between countries through factor analyses. Another way to study equivalence is
through examination of the nomological network. Quite a few studies have reported
correlations between the EPQ and context variables, both within and between
countries, but, again, no study has compared correlations at both levels. Therefore,
the second question addresses the functional equivalence of the EPQ at individual
and country level.
Method
Data Sets
Studies included in the secondary analyses were found by searching PsycLit
(now called PsycInfo), using the keywords EPQ and Eysenck. All volumes of the
 Eysenck Personality Questionnaire 22

journal Personality and Individual Differences were also searched. Further studies
were found through "snowballing" on the basis of references in literature already
identified. Studies were included if the averages on all four scales of the EPQ were
available.2 Only data on the 1975 version of the EPQ were used; EPQ-R, EPI and
JEPQ data were not included because of lack of comparability with the EPQ and an
insufficient number of relevant studies. As our aim was to compare data on normal
adult populations, studies with children, clinical samples or specific groups that
presumably show extreme scores on (some of) the scales, such as alcoholics and
gamblers, were not included in the study. Each separate group of respondents in the
same report, for which data were reported (e.g., men and women), was considered
as a separate sample. This procedure resulted in 153 studies, which provided data
for 333 separate samples. The data set contained studies in 38 countries with a total
of 68,374 respondents. Source, year of publication, composition of the sample
(general or students), sex, means and standard deviations of the four EPQ scales,
intercorrelations between the scales, and reliabilities of the scales were recorded, in
so far as they were available. This data set is called the total data set.
A second data set was used for the multilevel analyses, hereafter called the
multilevel data set. This set consisted of all studies from the first data set that
provided data on intercorrelations of the EPQ scales. Fifty-one (33.3%) studies with a
total of 25,922 respondents were found to meet these requirements, including 96
separate samples, from 24 countries.
Many studies had to be dropped because insufficient information was
provided by the author(s). For example, it was not always clear which version of the
EPQ was used. EPQ–R (revised) and the EPQ-RSS (short version) were regularly
referred to as EPQ. Additionally, not all required information could be found in each
study. In some cases the number of women and men in the sample was not
specified. Furthermore, some researchers did not fully describe the cultural
composition of their samples.
Both data sets contained more male samples: 45.6% of the samples in the
total data set and 49.0% in the multilevel data set were male. In the total data set
17.7% and in the multilevel data set 7.3% of the studies did not report data on gender
composition.
The age of the respondents ranged from 15 to 70.1 years for the total data set
and from 15 to 60 for the multilevel data set. The distribution of age was positively
skewed in both data sets. Mean ages were 27.46 (SD = 9.30) and 27.01 (SD = 8.89),
respectively. The value of the mode (21.00 in both data sets) reflects the
overrepresentation of students in EPQ research: 28.8% and 28.1% respectively of
 Eysenck Personality Questionnaire 23

the samples consisted solely of students. Data on the subjects’ age were missing for
39.6% and 14.6% of the studies respectively.
The publication dates of studies ranged from 1977 to 1997. There were no
earlier reports as the EPQ was only published in 1975. The mode was 1984 and the
median 1985 for both data sets.
A practical problem in comparing means on the EPQ scales across countries
lies in the fact that the scales as used do not contain equal numbers of items. As a
scale score is a sum of item scores, scale means had to be adjusted by multiplication
with correction factors (see Barrett & Eysenck, 1984).3 This correction procedure
assumes that the deleted items had average endorsement values. Reliabilities were
corrected to what they would be if scales would all have the same length as the
English versions, using the Spearman-Brown formula (Lord & Novick, 1968). The
number of items in the four scales ranged from 16 to 25 in EPQ-P, from 16 to 23 in
EPQ-E, from 18 to 23 in EPQ-N and from 19 to 24 in EPQ-L. Sample sizes, corrected
scale means and corrected scale reliabilities for all countries are shown in Table 1.
The means across the countries varied considerably, although means on the EPQ-P
were generally the lowest and means on the EPQ-E the highest. Reliabilities of the
EPQ-P tended to be considerably lower than reliabilities of the other scales.
Multilevel Factor Analysis
First, multilevel analyses were performed according to the adaptation of Van
de Vijver and Poortinga (2001) of the Muthén strategy (1991, 1994). The procedure
was as follows:
1. Exploratory factor analyses were carried out on the total data set to gain
information concerning the factor structure. Scores (per country) on the four
scales served as input. It would have been preferable to analyze data at item
level, because of the larger level of detail in the analyses. However, data at item
level were insufficiently available.
2. The pooled within-country correlation matrix (individual level) was computed,
based on the intercorrelations within each country. This matrix was factor
analyzed. The between-country correlation matrix (based on country-level
scores) was computed by averaging correlations per country (weighted by their
sample size). This country-level matrix was also factor analyzed.
3. To verify that the pooled-within structure applied to all countries, the factor
loadings derived from the pooled within-country matrix were compared with the
factor structures for each of the separate countries. The factorial agreement of
the pooled-within structure and each of the countries was evaluated.
 Eysenck Personality Questionnaire 24

4. Structures found in the pooled-within matrix and in the between matrices were
target rotated and Tucker's congruence coefficients (Tucker, 1951, p.43) per
scale and per factor were computed to evaluate their correspondence. Values
higher than .95 can be considered to indicate factorial similarity, whereas values
lower than .90 are assumed to point to non-negligible differences in factorial
structure (Van de Vijver & Poortinga, 1994).
5. Steps 2 to 4 were repeated in four subsequent analyses using only three scales,
successively leaving out the EPQ-P, EPQ-E, EPQ-N, and EPQ-L.
It should be noted that the main interest here is to compare factor structures at
individual and country level. The above-described analyses do not focus on the
dimensionality of the EPQ at either level. This implies that neither the size nor the
sign of factor loadings were of interest. Structural equivalence would be challenged
by a dissimilarity of structure, whatever the dimensionality found in the factor
analysis.
Lower bounds of congruence coefficients (i.e., threshold values below which
factors are taken to be dissimilar) were estimated using a bootstrap procedure
proposed by Chan et al. (1999). This bootstrap procedure allows an estimate of the
standard error of variable congruence coefficients; it evaluates the similarity of a
variable, comparing the loadings of that variable across factors (analogous to a factor
congruence coefficient which evaluates the agreement of the loadings of a factor
across variables). A raw data matrix was generated which yielded correlations equal
to the pooled-within correlations. From this data matrix 1000 samples were drawn.
Each sample consisted of 270 observations (being the original average sample size),
drawn with replacement from the original data matrix. The distribution of the
congruence coefficients for a variable was used to determine the lower bound (alpha
= .05).
Nomological Network Relationships
As no data at the individual level were available, functional equivalence at the
two levels was established by comparing previously reported individual level
correlations between the EPQ scales and other variables with correlations between
the EPQ scales and a number of country variables. The latter were drawn from
several sources.
Ecosocial factors. Georgas and Berry (1995) combined several ecological
and social indicators to six categories or factors. The Ecological factor included
highest and lowest average temperature and highest monthly level of precipitation.
The Economical factor included GNP per capita, daily calorie intake as a percentage
of the recommended amount, consumption of commercial energy per capita,
 Eysenck Personality Questionnaire 25

percentage of population working in agriculture, percentage of population working in

industry, percentage of population working in services, and electricity consumption
per capita. The third factor concerned the Educational system, and included the
teacher-pupil ratio in the first level, the gross enrollment in the first, the second and
the third level, and the percentage of adult illiterates. The fourth factor regarded the
means of Mass communication, reflected in the number of public and private
telephones per 100 inhabitants, newspapers (circulation per 1,000 inhabitants),
televisions (number of receivers per 1,000 inhabitants) and radios (number of
receivers per 1,000 inhabitants). The fifth factor, Population, included infant mortality
rates, life expectancy at birth, crude death rate, crude birth rate and rate of
population increase. Where indicators were not available for a country in our data
set, these were added by using the Demographic Yearbook 1987 (United Nations,
1990a), the UNESCO Statistical Yearbook 1990 (United Nations, 1990c), and the
Historical Climatology Series (National Climatic Data Center, 1991). An overall score
on Affluence (Georgas, Van de Vijver & Berry, 2001) was calculated by factor
analyzing all variables used for calculation of the five ecosocial factors. The resulting
factor scores on the single factor were used as a measure of Affluence. Data on the
five ecosocial factors and Affluence were available for all 38 countries.
Additional economics variables. The Gini index expresses the degree of
income inequality in a society. Indices for 29 countries (ranging from 1987 to 1995)
were collected from the World Development Report (World Bank, 1999). The Human
Development Index (United Nations, 1990b) was available for 36 countries. This
index measures development in three areas (life expectancy, adult literacy rate, and
Gross Domestic Product per Capita) in relation to other countries.
Sociopolitical factors. Five variables were selected to denote the social and
political atmosphere in countries. Humana (1986) collected data from several United
Nations instruments and constructed an index for rights and freedoms in 40
categories, the Human Rights Index (34 countries). Indices for Political Rights and for
Civil Liberties in the year 1984 to 1985 were available for 35 countries (Gastil, 1985).
Stability of Democracy (Inglehart, 1997) refers to the number of years of continuous
democracy (24 countries). Vanhanen’s (1997) Index of Democratization over the year
1980 is a weighted combination of two indicators of dimensions of democracy,
namely Competition (smaller parties’ share of the votes) and Participation
(percentage of total population who voted in the election).
Death rates. Death rates per 100,000 inhabitants by homicide, suicide, and
liver cirrhosis (as an indicator of alcoholism) were collected (following Lynn & Martin,
1995) using the Demographic Yearbook 1987 (United Nations, 1990a). Data were
 Eysenck Personality Questionnaire 26

available for 31 countries on homicide and suicide rates, and for 30 countries on liver
cirrhosis resulting in death. Further, as a measure of political instability the data by
Taylor and Jodice (1983) on death due to political violence between 1948 and 1977
were used. This measure, available for all countries, was corrected for population
size.
Bribe and corruption. Subjective data on bribery for 1999, obtained through
polls in emerging market countries, were provided by Transparency International
(Pope, 1999). This Bribe Payers Index was available for 14 countries. The Corruption
Perceptions Index from the same source indicates business people's opinion on
corruption in a large amount of countries in the period 1988-1992 (30 countries).
Religiosity. An analysis was done on a part of the 1990-1991 World Values
Survey (Inglehart, 1993, 1997). This study included 47,871 respondents from 39
countries. It provided data on a large range of topics related to religion, like the
meaning of life, religious services, the role of churches and praying. Six items on the
experience and practice of religion were selected from a larger number of indices to
form a scale of religiosity. Examples of items are: “Do you find that you get comfort
and strength from religion?” “How important is God in your life?” and “How often do
you pray to God outside of religious services?” The six items yielded a one-factor
solution in factor analysis at aggregated (country) level, with an eigenvalue of 5.39
(89.8% of variance explained). Cronbach’s alpha was .83. These data on religion
were available for 22 countries.
Anxiety. Mean scores on the question ‘How often do you feel nervous or
tense at work?’ (Hofstede, 1976) as published by Lynn (1981) were used for 23
countries.
Hofstede’s measures. Data for 23 countries were available on Hofstede’s
(1980) Individualism, Power Distance, Uncertainty Avoidance and Masculinity.
Subjective Well-being. A measure for subjective well-being was derived from
Diener, Diener, and Diener (1995). This value combines scores from several surveys
and was available for 30 countries.
Schwartz’s Values. Schwartz (1994) provided data on values in 22 of the
countries in the present study. Factor analysis by Georgas et al. (2001) yielded two
bipolar orthogonal factors, labelled Autonomy and Hierarchy. Scores on these two
dimensions were used as indicators.
Pace of Life. Levine and Norenzayan (1999) studied the pace of life in 31
countries. An index was established from three variables: Average walking speed
downtown, the speed with which postal clerks completed a simple request, and the
accuracy of public clocks. The Pace of Life index was reported to correlate positively
 Eysenck Personality Questionnaire 27

with GNP (r = .74; p < .01). Nineteen of the countries overlapped with countries in the
present study.
Results
Multilevel Analysis
Structural equivalence at the individual and the country level was established
first by performing multilevel analyses following the five steps described before. First
all four scales were factor analyzed. Across all countries (N = 24) a two-dimensional
structure was found. In the second step the pooled within-country correlation matrix
and the between-country correlation matrix were calculated. Table 2 shows both the
within-countries and the between-country correlation matrices. The third step
involved checking the factorial agreement of each country by comparing the factor
solution from the overall pooled-within correlation matrix with the factor solution of the
correlation matrix of the country means. It turned out that four countries did not meet
the requirements of an agreement coefficient > .90. These countries were China,
India, Japan, and Uganda. Sample characteristics in these countries did not distinctly
differ from the other countries. Although the data for China, Japan, and Uganda were
based on only one study, this was also the case for 13 of the other countries. Target
rotation and calculation of the congruence coefficients took place in the fourth step.
Coefficients for the total set were .96 and .78; after deleting these countries from this
analysis, Tucker’s coefficient was .99 for the first factor, but only .88 for the second
factor (Table 3). Thus, the first factor can be considered equivalent at individual level
and country level, whereas the second factor did not meet the requirements.
In order to examine the influence each of the four scales exerts on the
equivalence at individual and country level, four further analyses were performed,
each time omitting one scale (fifth step). When the EPQ-P was omitted Israel showed
a Tucker’s coefficient below .90 on the second factor (indicating poor equivalence
between the Israel matrix and the pooled within-country correlation matrix) and was
removed. Congruence coefficients (after target rotation) reached values of 1.00 and .
99 for the two-factor solution. The deletion of the EPQ-E resulted in coefficients of .
64 and .77 respectively, and the deletion of EPQ-N led to coefficient values of .90
and .89 respectively. In both these cases all countries showed sufficient agreement
(i.e., Tucker's coefficient > .90) with the factors based on the corresponding within-
country matrix. Finally, when leaving out the EPQ-L, China, Japan, Uganda, and
Yugoslavia showed a less than adequate fit and were removed. After extraction of,
again, two factors, agreement coefficients (after target rotation) of 1.00 and 1.00
were found (before deletion of the four countries Tucker’s coefficients were .99 and .
97).
 Eysenck Personality Questionnaire 28

These findings suggest that EPQ-E and EPQ-N define a single bipolar factor
that shows good structural equivalence within and across countries. EPQ-P and
EPQ-L, however, are more problematic. Elimination of either of the two is sufficient to
establish the functional equivalence of the three remaining scales.
In order to evaluate the equivalence of the four variables within and across
countries, congruence coefficients for the four variables were examined using the
Chan et al. (1999) procedure. Congruence coefficients for the EPQ scales were .54
(EPQ-P), 1.00 (EPQ-E), .96 (EPQ-N), and .93 (EPQ-L). Bootstrapping (with an alpha
level of .05) resulted in critical values of .91, .90, .96, and .95, for the four scales,
respectively. An observed coefficient smaller than the critical value points to
incongruence (Chan et al., 1999). Applied to the present data, it can be concluded
that both the EPQ-E and EPQ-N are congruent within and across countries, while the
EPQ-P and EPQ-L are incongruent. This result is in agreement with the findings from
the multilevel exploratory factor analyses.
Nomological Network Relationships
To study functional equivalence, within-country correlations were collected
from previously published studies and compared with between-country correlations
calculated from the total data set. Table 4 shows the nature of correlations found
between the EPQ scales and context variables within and between countries.
Because of relatively small sample sizes, correlations at country level hardly ever
reach significance. Therefore, the patterning of the correlations is more informative
than their level of significance. For the Lie scale it was established that correlations
with affluence or socioeconomic status are negative at both levels. Subjective well-
being appeared to be related to the Lie scale differently within and between
countries. However, the reasons for these differences between countries are still
unclear. Therefore, further correlations at country level were examined.
Correlations between the EPQ scales and the country characteristics are
presented in Table 5. It should be noted that findings in the preceding section cast
doubt on the exact meaning of some of the EPQ scales at country level. Also, sample
sizes are relatively small because of limited overlap between available data, reaching
a maximum of 38. A striking finding is the large number of significant correlations with
EPQ-L. High positive correlations were found between the EPQ-L and the Ecology (r
= .69; p < .01) and Population (r = .51; p < .01) factors. Extremely hot and humid
(tropical) countries, and countries with high birth and death rates and low life
expectancy score higher on social desirability. Strong negative correlations were
found with the ecosocial factors Economy (r = -.64; p < .01), Education (r = -.63; p < .
01), Communication (r = -.70; p < .01), and with Affluence (r = -.69; p < .01). Strong
 Eysenck Personality Questionnaire 29

negative correlations were also found for GNP and the Human Development Index.
Apparently, richer countries score lower on the EPQ-L. An identical pattern could be
seen for all sociopolitical variables: Lower social desirability scores were related to a
higher level of democracy and more observance of citizens' rights. As for death rates,
a significant negative correlation was found between the EPQ-L and the number of
deaths by suicide (r = -.45; p < .05). The occurrence of bribery and corruption went
together with high social desirability. A negative relation was found between the
EPQ-L and Individualism (r = -.68; p < .01). This is not surprising as individualism is
known to correlate with GNP, a finding confirmed in the present study (r = .66; p < .
01). Power distance was positively correlated with EPQ-L. Diener’s Subjective Well-
Being, also known to be related to GNP, correlates negatively with the EPQ-L.
Finally, the Pace of Life index correlated positively with the EPQ-L, and negatively
with GNP (r = -.55; p < .05). This result is in line with Levine and Norenzayan (1999),
who reported a negative correlation between Pace of Life and GNP.
The EPQ-P was significantly correlated with bribery. The often reported
negative correlation between EPQ-P and religiosity in individual level studies was not
replicated at country level (e.g., Francis, 1992; Heaven, 1990; Kay, 1981; Lewis &
Maltby, 1995; Svensen et al., 1992). Also, more masculine countries displayed higher
scores on psychoticism. Subjective well-being was negatively correlated with EPQ-P.
EPQ-E was correlated significantly with death resulting from political violence, the
Bribe Payers Index and religiosity. EPQ-N correlated with masculinity. The two
factors representing Schwartz’ values did not show significant correlations with any of
the EPQ scales.
Thus, most significant correlations were found with the EPQ-L. In general,
these correlations involve GNP related variables. To explore what other factors exert
influence, partial correlations for all country variables with the EPQ scales were
calculated, controlling for Affluence. All significant correlations with EPQ-L
disappeared, except for the positive correlation with the Ecological factor (r = .39; p <
.05), the negative correlations with the Population factor (r = -.40; p < .05) and
suicide rate (r = -.39; p < .05). Incidentally significant partial correlations with the
other EPQ scales were also observed. However, we did not find any consistent
interpretations for these.
Finally, the reliabilities of EPQ-P, EPQ-E, and EPQ-N were correlated with
the country scores on EPQ-L and with GNP. A factor analysis of EPQ-P, EPQ-E, and
EPQ-N reliabilities at country level yielded a one-factor solution with an eigenvalue
1.94, explaining 65.0% of the variance. Factor scores on this factor showed a
correlation of -.47 (p < .01) with the mean country score on EPQ-L, indicating that
 Eysenck Personality Questionnaire 30

reliabilities of the PEN scales are lower when scores on the Lie scale are higher.
Further, correlations between this reliability factor and GNP (r = .45; p < .05), and
both Affluence (r = .68) and the Education Factor (r = .61) were significant (p < .01).
This finding implies that social desirability is indeed related to reliability of the other
EPQ scales, and that on average reliabilities are lower in less affluent countries.
Discussion
Two questions were examined. The first concerned the multilevel structure of
the Eysenck Personality Questionnaire. Intercorrelations between the EPQ scales
were compared within and between countries. First, in some countries (China, India,
Japan, and Uganda) the correlations of the four EPQ scales differed from those
found elsewhere. Whether this is due to translations, sample characteristics,
differences in administration, or differences in construct is not clear. Second,
whereas the constructs of extraversion and neuroticism appear to have the same
psychological meaning within and across countries, this does not hold for
psychoticism and social desirability. Hence, the psychological meaning of differences
in score levels is not equivalent at individual and country level, at least not for
psychoticism and social desirability.
It may be argued that carrying out a factor analysis with only four input
variables (the four EPQ scale scores) is debatable, as the number of scales is small
and the underlying scales are assumed not to be strongly related. However, the aim
of the factor analysis was not the examination of the dimensionality of the EPQ within
and/or between countries, but the establishment of the (dis)similarity between factor
structures at both levels. For this purpose a factor analysis with four variables can be
regarded as acceptable. Also, the results from the bootstrap procedure can be
considered a confirmation of the findings.
The second research question involved the relationship of the EPQ scales
and context variables at individual and country level. It should be noted that the
nature of the results does not permit any statements about the direction of causality.
The clearest results were correlations with the Lie scale. At both levels more
affluence was accompanied by less social desirability. At the individual level
subjective well-being seemed to be positively related to social desirability or
conformity, while for countries the opposite was true. Correlations across countries
were further explored.
The main finding was the strong negative correlation of social desirability with
Gross National Product and related economic variables. The joint underlying factor
here is affluence. The negative correlations between the Lie scale and the factors
Education and Communication, Hofstede’s Individualism, and Subjective Well-Being
 Eysenck Personality Questionnaire 31

also are in agreement with this finding. All these variables are significantly correlated
with GNP and directly or indirectly reflect GNP or other wealth-related indices. This
strong social desirability effect can be interpreted in two ways, in line with two
existing models. First, it can be considered as due to method bias. This form of bias
refers to artefacts (for example instrument or administration characteristics)
influencing the results of an entire instrument (e.g., Van de Vijver & Leung, 1997).
According to this explanation differences in social desirability should be considered
as a kind of response bias. This explanation is supported by the finding that a higher
level of social desirability is related to lower reliabilities of the other three EPQ scales.
However, this finding is not in line with findings by Grimm and Church (1999) who
reported that cross-cultural differences in means on personality measures were ‘only
modestly’ confounded by response biases.
Second, the results can be interpreted as reflecting differences in social-
psychological functioning as a consequence of ecocultural conditions, such as GNP.
In this case, there is a psychological relationship between affluence and social
desirability. A sociocultural explanation for differences in social-psychological
functioning can be considered as well. Ross and Mirowsky (1984) state that less
powerful groups are more prone to socially desirable responding, less powerful
groups are often less affluent groups. People from these groups are forced to behave
according to social norms because they depend on the approval of other people. This
would suggest a relation at country level between social desirability and power-
related variables. At an individual level social desirability could easily be a measure
of ‘social naïveté’ or conformity (see Eysenck & Eysenck, 1975; Furnham, 1986;
Ones et al., 1996), social desirability being a personality characteristic. However, the
present findings could not support a relation between social desirability and power-
related variables, like Power Distance or Hierarchy, after controlling for GNP. More
research is needed to further explore which of the two interpretations of country
differences in social desirability, method bias or conformity, is more likely to be true.
Due to the nature of the data, we could not establish to what extent bias in
individual items has contributed to inequivalence. Although it has to be assumed that
item bias is present, it probably plays a small part in the correlational relationships
analysed in the present study. The basis for this assertion is the high agreement
across countries.
It should be noted that Lynn and Martin's (1995) findings were not replicated,
as no significant correlations were found between extraversion and any of the death
rate indices. However, suicide rate was related to the Lie scale. Also, Lynn’s
hypotheses about the association between political, social, and economic instability,
 Eysenck Personality Questionnaire 32

and climate (Lynn, 1981), and neuroticism across countries were not supported. The
same applies to his claims about the relationship between extraversion and income.
Most likely, sample differences explain the failure to find similar results, as the
countries and the studies from which scores were derived only partially overlap.
Moreover, Lynn's results were based on various questionnaires, such as the EPI,
MPI, and EPQ. Except for the 1995 study, the sample size in Lynn's studies was
considerably lower than in the present study. Furthermore, Lynn did not evaluate
relations at the individual level.
In summary, two EPQ scales (Extraversion and Neuroticism) were
convincingly equivalent at an individual and country level. Very few significant
correlations with country variables could be found for these scales. For the two other
scales (Psychoticism and Lie) aggregating individual scores to country level seems to
affect the psychological meaning of the scores. For the Lie scale significant
correlations with many country level variables were found; all of these seem to reflect
differences in affluence. Thus, it was established quite convincingly that social
desirability scores are higher in countries that are poorer, less educated, less
egalitarian and more collectivistic.
 Eysenck Personality Questionnaire 33

Footnotes
1
In the present article the term individual level refers to within-country scores
or matrices and the term country level refers to between-country scores or matrices
(see for similar terminology, e.g., Hofstede, 1991; Leung, 1989; Leung & Bond, 1989;
Van de Vijver & Leung, in press; Van de Vijver & Poortinga, 2001).]
2
In some cases, however, the data on the L scale or the P scale were
missing.
3
The 8 most heavily corrected countries were Bangladesh, Brazil, Iceland,
Nigeria, Norway, Spain, Sri Lanka, and Russia. In later analyses these countries did
not show a deviant pattern of agreement with the overall structure.
 Eysenck Personality Questionnaire 34

References
A list of references included in the meta-analysis can be found in Appendix A.
Arrindell, W. A., Hatzichristou, C., Wensink, J., Rosenberg, E., Van Twillert, B.,
Stedema, J., & Meijer, D. (1997). Dimensions of national culture as predictors
of cross-national differences in subjective well-being. Personality and
Individual differences, 23, 37-53.
Barrett, P., & Eysenck, S. (1984). The assessment of personality factors across 25
countries. Personality and Individual Differences, 5, 615-632.
Barrett, P. T., Petrides, K. V., Eysenck, S. B. G., & Eysenck, H. J. (1998). The
Eysenck Personality Questionnaire: An examination of the factorial similarity
of P, E, N, and L across 34 countries. Personality and Individual Differences,
25, 805-819.
Bijnen, E. J., & Poortinga, Y. H. (1988). The questionable value of cross-cultural
comparisons with the Eysenck Personality Questionnaire. Journal of Cross-
Cultural Psychology, 19, 193-202.
Bijnen, E. J., Van der Net, T. Z. J., & Poortinga, Y. H. (1986). On cross-cultural
comparative studies with the Eysenck Personality Questionnaire. Journal of
Cross-Cultural Psychology, 17, 3-16.
Bond, R., & Smith, P. B. (1996). Culture and conformity: A meta-analysis of studies
using Asch’s (1952b, 1956) Line Judgment Task. Psychological Bulletin, 119,
111-137.
Caird, D. (1987). Religiosity and personality: Are mystics introverted, neurotic, or
psychotic? British Journal of Social Psychology, 26, 345-346.
Chan W., Ho, R. M., Leung, K., Chan, D. K-S, & Yung, Y.-F. (1999). An alternative
method for evaluating congruence coefficients with Procrustes rotation: A
bootstrap procedure. Psychological Methods, 4, 378-402.
Crandall, V. C., Crandall, V. J., & Katkovsky, W. (1965). A children’s social
desirability questionnaire. Journal of Consulting Psychology, 29, 27-36.
DeNeve, K. M., & Cooper, H. (1998). The happy personality: A meta-analysis of 137
personality traits and subjective well-being. Psychological Bulletin, 124, 197-
229.
Diener, E., Diener, M., & Diener, C. (1995). Factors predicting subjective well-being
of nations. Journal of Personality and Social Psychology, 69, 851-864.
Edwards, D., & Riordan, S. (1994). Learned resourcefulness in black and white
South African university students. Journal of Social Psychology, 134, 665-
675.
 Eysenck Personality Questionnaire 35

Eysenck, H. J., & Eysenck, S. B. G. (1975). Manual of the Eysenck Personality

Questionnaire. London: Hodder and Stoughton.
Eysenck, H. J., & Eysenck, S. B. G. (1976). Psychoticism as a dimension of
personality (Ch. 11). London: Hodder and Stoughton.
Eysenck, H. J., & Eysenck, S. B. G. (1983). Recent advances in the cross-cultural
study of personality. In J. N. Butcher & C. D. Spielberger (Eds.), Advances in
personality assessment (Vol.2) (pp. 41-69). Hillsdale, NJ: Erlbaum.
Eysenck, S. B. G., Barrett, P., Spielberger, C., Evans, F. J., & Eysenck, H. J. (1986).
Cross-cultural comparisons of personality dimensions: England and America.
Personality and individual differences, 7, 209-214.
Eysenck, S. B. G., & Haraldsson, E. (1983). National differences in personality:
Iceland and England. Psychological Reports, 53, 999-1003.
Fisher, G. (1967). The performance of male prisoners on the Marlowe-Crowne social
desirability scale: II. Differences as a function of race and crime. Journal of
Clinical Psychology, 23, 473-475.
Francis, L. J. (1985). Personality and religion: theory and measurement. In L. B.
Brown (Ed.), Advances in the psychology of religion. International series in
experimental social psychology (Vol. 11, pp. 171-184). Oxford, UK:
Pergamon.
Francis, L. J. (1992). Is psychoticism really a dimension of personality fundamental to
religiosity? Personality and Individual Differences, 13, 645-652.
Francis, L. J., Lewis, J. M., Brown, L. B., Philipchalk, R., & Lester, D. (1995).
Personality and religion among undergraduate students in the United
Kingdom, United States, Australia and Canada. Journal of Psychology and
Christianity, 14, 250-262.
Francis, L. J., Pearson, P. R., Carter, M., & Kay, W. K. (1981). Are introverts more
religious? British Journal of Social Psychology, 20, 101-104.
Furnham, A. (1986). Response bias, social desirability and dissimulation. Personality
and Individual Differences, 7, 385-400.
Gastil, R. D. (Ed.). (1985). Freedom in the world: Political rights and civil liberties,
1984-1985. Westport, CT: Greenfield.
Georgas, J., & Berry, J. W. (1995). An ecocultural taxonomy for cross-cultural
psychology. Cross-Cultural Research, 29, 121-157.
Georgas, J., Van de Vijver, F. J. R., & Berry, J. W. (2001). Ecosocial indicators and
psychological variables in cross-cultural research. Manuscript submitted for
publication.
 Eysenck Personality Questionnaire 36

Gough, H. G., & Heilbrun, A. B., Jr. (1965). Adjective Check List manual. Palo Alto,
CA: Consulting Psychologists Press.
Grimm, S. D., & Church, A. T. (1999). A cross-cultural study of response biases in
personality measures. Journal of Research in Personality, 33, 415-441.
Heaven, P. C. L. (1990). Religious values and personality dimensions. Personality
and Individual Differences, 11, 953-956.
Hofstede, G. (1976). Nationality and organisational stress. Brussels: European
Institute for Advanced Studies in Management.
Hofstede, G. (1980). Culture’s consequences. Beverly Hills, CA: Sage.
Hofstede, G. (1991). Empirical models of cultural differences. In N. Bleichrodt & P. J.
D. Drenth (Eds.), Contemporary issues in cross-cultural psychology (pp.4-20).
Lisse, The Netherlands: Swets & Zeitlinger.
Humana, C. (1986). World human rights guide. London: The Economist Publications.
Inglehart, R. (1993). World values survey 1990-1991. WVS Program. J.D. Systems,
S.L. ASEP S.A.
Inglehart, R. (1997). Modernization and postmodernization. Cultural, economic, and
political change in 43 countries. Princeton, NJ: Princeton University Press.
Johnson, T., & Van de Vijver, F. J. R. (2000). Social desirability from a cross-cultural
research perspective. Manuscript in preparation.
Kaiser, H. F., Hunka, S., & Bianchini, J. C. (1971). Relating factors between studies
based upon different individuals. Multivariate Behavioral Research, 6, 409-
422.
Kay, W. K. (1981). Psychoticism and attitude to religion. Personality and Individual
Differences, 2, 249-252.
Klassen, D., Hornstra, R. K., & Anderson, P. B. (1975). Influence of social desirability
on symptom and mood reporting in a community survey. Journal of
Consulting and Clinical Psychology, 43, 448-452.
Lester, D. (1988). National character and rates of personal violence (suicide and
homicide). Personality and Individual Differences, 9, 423.
Lester, D. (2000). National differences in neuroticism and extraversion. Personality
and Individual Differences, 28, 35-39.
Leung, K. (1989). Cross-cultural differences: Individual-level vs. culture-level
analysis. International Journal of Psychology, 24, 703-719.
Leung, K., & Bond, M. H. (1989). On the empirical identification of dimensions for
cross-cultural comparisons. Journal of Cross-Cultural Psychology, 20, 133-
151.
 Eysenck Personality Questionnaire 37

Levine, R. V., & Norenzayan, A. (1999). The pace of life in 31 countries. Journal of
Cross-Cultural Psychology, 30, 178-205.
Lewis, C. A., & Maltby, J. (1995). Religiosity and personality among U.S. adults.
Personality and Individual Differences, 18, 293-295.
Lolas, F., Gomez, A., & Suarez, L. (1991). EPQ-R and suicide attempt: The
relevance of psychoticism. Personality and Individual Differences, 12, 899-
902.
Lord, F. M., & Novick, M. R. (1968). Statistical theories of mental test scores.
Reading, MA: Addison-Wesley.
Lynn, R. (1971). Personality and national character. Oxford, UK: Pergamon.
Lynn, R. (1981). Cross-cultural differences in neuroticism, extraversion and
psychoticism. In R. Lynn (Ed.), Dimensions of personality. Papers in honour
of H. J. Eysenck (pp. 263-286). Oxford, UK: Pergamon.
Lynn, R., & Hampson, S. L. (1975). National differences in extraversion and
neuroticism. British Journal of Social and Clinical Psychology, 14, 223-240.
Lynn, R., & Hampson, S. L. (1977). Fluctuations in national levels of neuroticism and
extraversion. British Journal of Social and Clinical Psychology, 16, 131-138.
Lynn, R., & Martin, T. (1995). National differences for thirty-seven nations in
extraversion, neuroticism, psychoticism and economic, demographic and
other correlates. Personality and Individual Differences, 19, 403-406.
McCrae, R. R., & Costa, P. T., Jr. (1983). Social desirability scales: More substance
than style. Journal of Consulting and Clinical Psychology, 51, 882-888.
Muthén, B. O. (1991). Multilevel factor analysis of class and student achievement
components. Journal of Educational Measurement, 28, 338-354.
Muthén, B. O. (1994). Multilevel covariance structure analysis. Sociological Methods
& Research, 22, 376-398.
Myers, D. G., & Diener, E. (1996). The pursuit of happiness. Scientific American,
274, 54-57.
Ones, D. S., Viswesvaran, C., & Reiss, A. D. (1996). Role of social desirability in
personality testing for personnel selection: The red herring. Journal of Applied
Psychology, 81, 660-679.
Poortinga, Y. H., & Van de Vijver, F. J. R. (2000). How different are personality traits
cross-culturally, and how are they different? [Review of the book The
importance of psychological traits: A cross-cultural study]. Contemporary
Psychology, 45, 89-91.
Pope, J. (Ed.). (1999). National integrity systems: The TI source book. Berlin:
Transparency International.
 Eysenck Personality Questionnaire 38

Pritchard, W. S., & Kay, D. L. C. (1993). Personality and smoking motivation of U.S.
smokers as measured by the State-Trait Personality Inventory, the Eysenck
Personality Questionnaire, and Spielberger’s Smoking Motivation
Questionnaire. Personality and Individual Differences, 14, 629-637.
Ross, C. E., & Mirowsky, J. (1984). Socially-desirable response and acquiescence in
a cross-cultural survey of mental health. Journal of Health and Social
Behavior, 25, 189-197.
Schuckit, M. A., Klein, J., Twitchell, G., & Smith, T. (1994). Personality test scores as
predictors of alcoholism almost a decade later. American Journal of
Psychiatry, 151,1038-1042.
Schwartz, S. H. (1994). Beyond individualism/collectivism: New cultural dimensions
of values. In U. Kim, H. C. Triandis, C. Kagitcibasi, S. C. Choi, & G. Yoon
(Eds.), Individualism and collectivism (pp. 85-119). Thousand Oaks, CA:
Sage.
Šipka, P. (1988). Personality correlates of faking personality tests: A cross-cultural
perspective. In D. H. Saklofske & S. B. G. Eysenck (Eds.), Individual
differences in children and adolescents: International perspectives (pp. 321-
338). London: Hodder and Stoughton.
Svensen, S. G., White, K. D., & Caird, D. (1992). Replications and resolutions:
Dualistic belief, personality, religiosity, and paranormal belief in Australian
students. Journal of Psychology, 126, 445-447.
Taylor, C. L., & Jodice, D. A. (1983). World handbook of political and social
indicators: Vol. 2. Political protest and government change. New Haven, CT:
Yale University Press.
Tucker, L. R. (1951). A method for synthesis of factor analysis studies (Personnel
Research Section Report No. 984). Washington, DC: Department of the
Army.
United Nations (1990a). United Nations demographic yearbook 1987. New York:
Author.
United Nations (1990b). Human development report. New York: Author.
United Nations (1990c). Unesco statistical yearbook 1990. New York: Author.
National Climatic Data Center (1991). Historical climatology series 6-4: Climates of
the world. Washington DC: US Department of Commerce.
Van de Vijver, F. J. R., & Leung, K. (1997). Methods and data analysis for cross-
cultural research. Newbury Park, CA: Sage.
Van de Vijver, F. J. R., & Leung, K. (in press). Personality in cultural context:
Methodological issues. Journal of Personality.
 Eysenck Personality Questionnaire 39

Van de Vijver, F. J. R., & Poortinga, Y. H. (1994). Methodological issues in cross-

cultural studies on parental rearing behavior and psychopathology. In C.
Perris, W. A. Arrindell & M. Eisemann (Eds.), Parenting and psychopathology
(pp. 173-197). Chichester, UK: Wiley.
Van de Vijver, F. J. R., & Poortinga, Y. H. (2001). Structural equivalence in multilevel
research. Manuscript submitted for publication.
Vanhanen, T. (1997). Prospects of democracy: A study of 172 countries. London:
Routledge.
Warnecke, R. B., Johnson, T. P., Chavez, N., Sudman, S., O’Rourke, D. P., Lacey,
L., & Horm, J. (1997). Improving question wording in surveys of culturally
diverse populations. Annals of Epidemiology 7, 334-342.
Williams, J. E., Satterwhite, R. C., & Saiz, J. L. (1998). The importance of
psychological traits. New York: Plenum.
World Bank (1999). World development report 1998/1999. Knowledge for
development. New York: Oxford University Press.
 Eysenck Personality Questionnaire 40

Table 1
Country Sample Sizes, Scale Means, and Reliabilities
N Scale averages Internal consistencies
Country Total data set Multilevel data Pb Ec Nd Le Pb Ec Nd Le
seta
Australia 1452 480 6.77 23.17 18.02 10.24 .74 .89 .90 .84
Bangladesh 1075 -- 4.25 19.05 12.28 19.14 .74 .82 .80 .81
Brazil 1396 1396 3.94 17.58 14.93 17.97 .66 .80 .77 .77
Bulgaria 1038 -- 4.17 18.60 14.96 15.11 .70 .82 .84 .77
Canada 1652 988 3.79 20.67 15.07 9.46 .60 .82 .85 .70
Chile 67 -- 1.97 22.73 8.20 -- -- -- -- --
China 2097 2097 8.28 13.01 14.08 18.83 -- -- -- --
Czechoslovakia 804 -- 5.23 13.83 13.70 12.42 -- -- -- --
Egypt 1330 -- 4.38 18.55 17.40 21.41 .53 .77 .80 .77
Finland 949 949 4.93 16.28 14.52 11.49 .68 .88 .84 .82
France 866 -- 4.72 11.81 11.32 -- -- -- -- --
Greece 1301 -- 5.47 20.38 18.34 16.61 .65 .82 .80 .82
Hong Kong 732 732 6.71 16.55 14.70 14.57 .63 .80 .83 .69
Hungary 962 -- 3.76 16.70 14.37 12.65 .65 .82 .84 .82
Iceland 1144 1144 3.52 19.18 13.89 10.53 .62 .82 .80 .72
India 2275 1294 6.22 20.07 14.44 15.17 .57 .74 .80 .75
Iran 624 -- 5.11 15.08 13.13 16.54 .76 .77 .82 .76
Ireland 2804 -- 4.65 18.85 13.17 9.72 .67 .81 .84 .68
Israel 2412 1050 4.35 22.02 9.08 16.34 .57 .78 .80 .80
Italy 2609 1824 5.71 17.46 16.45 16.88 .74 .83 .84 .76
Japan 258 258 6.94 17.17 17.55 10.56 .68 .80 .74 .67
Netherlands 1401 876 4.71 20.30 13.82 13.19 .70 .86 .87 .76
Nigeria 430 430 3.58 24.69 8.60 18.25 .52 .67 .78 .79
Norway 802 802 2.19 18.62 10.41 11.75 .80 .85 .84 .81
Poland 120 -- 8.06 17.01 14.53 10.95 -- -- -- --
 Eysenck Personality Questionnaire 41

Puerto Rico 1094 -- 4.41 21.00 14.16 17.03 .65 .79 .84 .84
Russia 1067 1067 3.49 16.12 17.87 14.16 .69 .84 .82 .83
Saudi-Arabia 600 -- 6.15 18.79 15.55 16.71 -- -- -- --
Singapore 994 994 4.35 17.41 13.03 16.32 .62 .79 .85 .79
Spain 2986 199 5.19 17.98 16.04 14.11 .70 .85 .85 .72
Sri Lanka 1027 1027 4.26 11.04 12.09 20.88 .62 .71 .81 .81
Sweden 126 -- 3.94 19.70 5.36 12.54 -- -- -- --
Uganda 1476 1476 6.04 19.63 15.49 13.61 .74 .72 .75 .77
United Kingdom 17725 2945 4.45 18.94 14.42 9.86 .67 .86 .85 .81
United States 4153 1279 3.67 20.83 13.78 11.54 .56 .82 .85 .76
West-Germany 2538 388 6.60 18.88 13.87 10.31 .63 .85 .85 .83
Yugoslavia 1430 971 7.07 18.31 13.85 15.90 .91 .82 .84 .82
Zimbabwe 2758 1256 5.86 18.48 15.37 14.64 .61 .76 .79 .78
Totalf 68574 25922 4.96 18.63 14.34 13.23 .66 .82 .83 .78
a
Dashes indicate no studies were suitable for the multilevel analyses
b
Psychoticism
c
Extraversion
d
Neuroticism
e
Lie
f
Means across all studies, weighted by their sample size
 Eysenck Personality Questionnaire 42

Table 2
Intercorrelations between EPQ scales for the Multilevel Data Set
a
Scale P E N L
Psychoticism (P) -- -.25 .37 -.02
Extraversion (E) -.00 -- -.35 -.39
Neuroticism (N) .15 -.17 -- -.35
Lie (L) -.27 -.06 -.21 --
a
Correlations above the diagonal represent the between-country correlations,
correlations below the diagonal represent the pooled within-country correlations
Note. Correlations are based on 24 countries
 Eysenck Personality Questionnaire 43

Table 3
Factor loadings (Varimax Rotated), Eigenvalues of Between and Within Correlation
Matrices, and Agreement Indices
Pa Eb Nc Ld Eigenvalues % Explained Tucker’s
Variance Congruence
Coefficient
Between (n = 20)
Factors
1 .66 -.42 .89 -.42 1.58 39.48 --
2 -.03 -.83 .04 .82 1.35 33.82 --
Within
1 .73 .21 .40 -.79 1.42 35.49 .99
2 -.05 .85 -.67 -.02 1.12 27.96 .88
RMSDe .33 .06 .15 .20
P omitted – Between (n = 23)
Factors
1 -- .89 -.01 -.84 1.51 50.27 --
2 -- -.30 .97 .42 1.18 39.41 --
P omitted - Within
1 -- -.07 -.68 .85 1.24 41.16 1.00
2 -- .91 -.48 -.23 1.07 35.59 .97
RMSDe -- .14 .09 .15
E omitted – Between (n = 24)
Factors
1 .93 -- .63 .04 1.52 50.74 --
2 .07 -- -.59 .94 .98 32.52 --
E omitted – Within
1 .86 -- .10 -.71 1.42 47.21 .64
2 -.05 -- .97 -.29 .86 28.72 .77
RMSDe .07 -- .50 .67
N omitted – Between (n = 24)
Factors
1 .02 -.77 -- .89 1.46 48.61 --
2 .96 -.40 -- -.17 1.02 34.08 --
 Eysenck Personality Questionnaire 44

N omitted – Within
1 .80 .02 -- -.79 1.27 42.40 .90
2 -.10 .99 -- -.13 1.00 33.40 .89
RMSDe .31 .10 -- .38
L omitted – Between (n = 20)
Factors
1 .92 -.11 .63-- 1.65 54.91 --
2 -.04 .95 -.48-- .75 24.90 --
L omitted - Within
1 .05 .87 -.62 -- 1.22 40.83 1.00
2 .91 .16 .47 -- 1.00 33.18 1.00
e
RMSD .08 .08 .03 --
Note. Two factors were drawn in each analysis to enable target rotation later
a
Psychoticism
b
Extraversion
c
Neuroticism
d
Lie
e
Square root of the mean squared difference per scale
 Eysenck Personality Questionnaire 45

Table 4
Sign and Strength of Correlations Found between EPQ scales and Context Variables
Within-Countries and Between-Country
P E N L
Wa Bb W B W B W B
Affluence n.s. n.s. + n.s. - n.s. - -
Suicidec + n.s. n.s. n.s. + n.s. n.s. -
Alcoholism ? n.s. + n.s. n.s. n.s. n.s. n.s.
Religiosity - n.s. n.s. + -+ n.s. + n.s.
Subjective ? - + n.s. - n.s. + -
well-being
a
Within-countries correlations, as found in literature (see sources)
b
Between-country correlations, as found in the present study
c
Number of suicide attempts and suicide ideation
Note. Nature of correlations is denoted by ‘’n.s.’’ (no significant correlation), ‘’-‘’ (a
negative correlation), ‘’+’’ (a positive correlation), or a combination of these (''-+''). A
lack of findings is denoted by ''?''.
Sources: Pritchard & Kay, 1993; Ross & Mirowsky, 1984; Lolas, Gomez, & Suarez,
1991; Schuckit, Klein, Twitchell, & Smith, 1994; Francis, 1992; Francis et al., 1981;
Caird, 1987; Francis, 1985; DeNeve & Cooper, 1998
 Eysenck Personality Questionnaire 46

Table 5
Correlations (Pearson) between Mean Scale Scores and Context Variables for the
Total Data set
Country variable Pa Eb Nc Ld
Ecosocial factors
Economy -.23 .10 -.04 -.64**
Ecology .07 .05 -.11 .69**
Communication -.15 .10 -.00 -.70**
Education -.09 -.09 .10 -.63**
Population .04 .18 -.06 .51**
Affluence -.13 -.03 .04 -.69**
Additional economics variables
Gross National Product -.19 .11 -.06 -.67**
Gini index -.11 .26 -.05 .29
Human Development Index -.06 -.14 .03 -.57**
Sociopolitical factors
Human Rights Index -.08 .21 -.03 -.49**
Political Rightse -.10 .12 .13 -.46**
Civil Libertiesf -.19 .24 .03 -.53**
Stability of Democracy -.20 .20 -.11 -.63**
Index of Democratization -.06 .28 -.08 -.48**
Death rates
Homicide -.10 .02 .07 .16
Suicide .12 -.13 -.07 -.45*
Liver Cirrhosis -.09 -.25 -.04 -.19
Deaths from Political Violenceg -.17 .35* -.29 .19
Bribe and corruption
Bribe Payers Indexh .58* -.73** .27 .73**
Corruption Perceptions Indexi .03 .05 .14 .71**
Religiosity -.34 .66** -.11 .04
Anxietyj .22 .04 .34 .06
Hofstede’s measures
Individualism -.11 .18 .05 -.68**
Masculinity .47* -.04 .63** -.17
Power Distance .26 -.39 .25 .58**
Uncertainty Avoidance .14 -.05 .23 .30
Subjective well-being -.41* .29 -.23 -.57**
Schwartz' values
Autonomy factor .07 -.11 .14 -.11
Hierarchy factor .24 .04 -.08 .28
Pace of lifek .23 .13 -.19 -.63**
a
Psychoticism
b
Extraversion
c
Neuroticism
d
Lie
efhijk
Original sign changed
g
Corrected for population size
*p < .05. **p < .01.