This is the html version of the file https://www.researchgate.net/profile/Jaap_Dronkers/publication/228676542_Ethnic_diversity_in_neighborhoods_and_individual_trust_of_immigrants_and_natives_A_replication_of_Putnam_(2007)_in_a_West-European_country/links/09e415092b10b1aeea000000.pdf.
Google automatically generates html versions of documents as we crawl the web.
These search terms have been highlighted: putnam ethnic
A decrease of social trust and capital by increased immigration and ethnic diversity
Page 1
Ethnic diversity in neighborhoods and individual trust of
immigrants and natives:
A replication of Putnam (2007) in a West-European country.
BRAM LANCEE AND JAAP DRONKERS
European University Institute
Paper presented at the International Conference on Theoretical Perspectives on Social Cohesion and
Social Capital, Royal Flemish Academy of Belgium for Science and the Arts, Brussels, Palace of
the Academy. May 15, 2008
Page 2
1
Ethnic diversity in neighborhoods and individual trust of immigrants and
natives:
A replication of Putnam (2007) in a West-European country.
BRAM LANCEE AND JAAP DRONKERS
*
Abstract. Putnam (2007) claims that in the short run ethnic diversity tends to reduce
solidarity and social capital: in ethnically diverse neighborhoods, residents of all
ethnicities tend to ‘hunker down’. Trust (even in one’s own ethnic group) is lower,
altruism and community cooperation is more rare, friends fewer. This study replicates
Putnam’s analysis for a West-European country. Furthermore, by including the
ethnicity of the respondent’s neighbors, a sub-neighborhood level measure of ethnic
diversity is added to the analyses. With data from the Netherlands (N=5,757), using
multi-level regression, we confirm Putnam’s claim and find that both for immigrants
and native residents 1) neighborhoods’ ethnic diversity reduces individual trust in
neighborhoods; 2) those with neighbors of a different ethnicity have less trust in
neighborhoods and neighbors 3) a substantial part of the effect of neighborhoods’
ethnic diversity on individual trust can be explained by the higher propensity of having
neighbors of a different ethnicity. We conclude that ethnic diversity can have a
negative effect on individual trust. However, we do not find these negative effects of
neighborhoods’ or neighbors ethnic diversity on inter-ethnic trust.
Key words: ethnic diversity, trust, social capital, neighborhoods, West-Europe.
1. Introduction
Putnam (2007) claims that in the short run, immigration and ethnic diversity tend to reduce
solidarity and social capital. He presents new evidence from the US that in ethnically diverse
neighborhoods residents of all ethnic groups tend to ‘hunker down’. Trust (even in one’s own ethnic
group) is lower, altruism and community cooperation is more rare, friends fewer. Since Putnam only
analyses data for the US, he proposes to test his claim in other countries. Giddens (2007), although
he acknowledges the tensions between diversity and the welfare state in Europe, believes that it is
impossible to apply the same analysis to Europe, because of a lack of data. In this paper, using data
from the Netherlands, we do apply an analysis analogous to that of Putnam in one of the EU
member states.1
Gesthuizen, Van der Meer and Scheepers (Forthcoming; see also Hooghe, Reeskens, Stolle, &
Trappers, 2006) also test Putnam’s thesis. They refute the claim that ethnic diversity results in less
social capital. However, both analyze ethnic diversity on the national level; using data from the
Euro barometer or the European Social Survey. Differences in ethnic diversity at a national level are
not the same as ethnic diversity in the neighborhood, as investigated by Putnam. As a consequence,
these indexes of diversity at the country level might be indicators of general social and political
processes relating not related with ethnic diversity (see for the dangers of ecological fallacy:
Snijders & Bosker, 1999). Therefore, their rejection of the Putnam findings with European data
might be fully wrong due to ecological fallacy. Effects of ethnic diversity on the country level rather
refer to the history, political environment and/or the media exposure of immigration related issues in
a country. Hence, effects on a national level are not the same as effects on the neighborhood level
(Murie & Musterd, 2004). Ethnic diversity in a country does for example not necessarily imply that
*
We would like to thank dr. Bart Bakker of the Dutch Statistical office for his help with the connection between the zip
code of the respondent and zip code level variables. Bram Lancee: bram.lancee@eui.eu; Jaap Dronkers:
jaap.dronkers@eui.eu
Page 3
2
an individual actually has inter-ethnic contacts. This may explain why Gesthuizen et al. and Hooghe
et al. do not find effects of ethnic diversity: the mechanism at hand is different. As opposed to ethnic
diversity in a country, in an ethnically diverse neighborhood it is much more likely that people of
different ethnicities actually meet. Face-to–face contact in turn affects sociological and
psychological processes such as building trust. Hence, choosing a different level of analysis may
imply exploring a different mechanism between ethnic diversity and social trust. It can therefore be
expected that results are different. As Murie and Mustard conclude (2004, p. 1441): ‘Understanding
the role of neighborhood in social exclusion involves attention to different levels of analysis and
different fault lines and to the resources that are produced within neighborhoods’. To further
investigate Putnam’s claim, an analysis on the neighborhood level is necessary.
The objective of this paper is twofold: 1) testing Putnam’s hypothesis in an European context,
by replicating his study as closely as possible in this European context; and 2) extending Putnam’s
analysis by adding ethnicity of the respondent’s neighbors as a lower level measure of ethnic
diversity, and thus explaining the relation between the neighborhood diversity and individual trust
by the ethnic diversity of concrete neighbors. We don’ t have additional theoretical claims about
trust and ethnic diversity, more profound than those of Putnam, but we would like to explain a
possible relation between neighborhood diversity and individual trust by a less abstract mechanism
(in our case the neighbors).
Our main question is “Can individual trust in neighbors and the neighborhood also in Europe
be explained by ethnic diversity in the neighborhood and that of the neighbors?” To answer this
question we make use of multi-level regression, which allows us to control for individual,
neighborhood, and municipality characteristics. As Gesthuizen et. al. (Forthcoming) and Hooghe et
al. (2006) also note, multi-level regression allows to disentangle the contextual and individual
characteristics in a more sophisticated matter than in the OLS regression used by Putnam.
2. Theory and hypotheses
As a result of continuing immigration, the population of modern societies is becoming more and
more ethnically heterogeneous. One of the major challenges that these countries face is how to deal
with ethnic diversity. Putnam claims that –at least in the short run- ethnic diversity in the
neighborhood reduces social solidarity and social capital.
Social capital implies that people well equipped with social resources – in the sense of their
social network and the resources of others they can call upon- succeed better in attaining their goals.
Hence, it is generally thought to be a true ‘capital’, in the sense that it yields positive returns (Portes,
2000). This holds for different outcomes, such as for example labor market performance
(Granovetter, 1973, 1995; Aguilera, 2002; Lin, 1999) or political participation (Fennema & Tillie,
1999). More specifically, also social capital within the neighborhood appears to be beneficial for
individuals, for example with respect to home-ownership (Brisson & Usher, 2007) or social
participation (Dekker, 2007). However, Alesina and La Ferrara (2000) find for the US that ethnic
diversity in the neighborhood negatively affects participation in social activities. It therefore seems
worthwhile to further explore the impact of ethnic diversity on social capital in the neighborhood.
There are many different forms or elements of social capital; consequently even more
definitions of the concept are being applied. As a result, Putnam (2007) chooses to adopt a ‘mean
and lean’ definition. A helpful approach when conceptualizing social capital is to distinguish
between structural and cognitive social capital (Poortinga, 2006; Lancee, Forthcoming). The
structural component refers to the ‘wires’ in the network: the frequency and intensity of links that
contribute to the exchange of resources. As opposed to cognitive social capital, structural social
capital involves a behavioral component. The cognitive component refers to the ‘nodes’ in a
network: attitudes and values such as perceptions of support, reciprocity and trust that contribute to
the exchange of resources.
The difference between bonding and bridging social capital might also be a useful
explanation of the relation between neighborhood diversity and individual trust. Social capital
Page 4
3
needed for bridging with other groups is more difficult to achieve and requires more investment
than social capital for bonding within ones own group (Burt, 2002; 2001). A neighborhood with a
high level of ethnic diversity requires more bridging social capital than a neighborhood with a low
level of ethnic diversity, but since this is more costly, bridging social capital is not sufficiently
available to maintain an average level of trust in neighbors and neighborhoods.
Social solidarity is characterized by the motive of promoting group goals in their own right,
provided the actor perceives positive attitudes from others towards himself (Wildegren, 1997;
Bourgeois & Friedkin, 2001). Most of the measures presented by Putnam (2007) -i.e. social trust
and solidarity- can be characterized as cognitive social capital. However, when analyzing social
capital in the neighborhood, the measurement benefits from including a behavioral component as
well. For example, Van der Laan Bouma-Doff (2007; see also Dagevos & Ode, 2003) finds that for
immigrants in the Netherlands ethnic concentration exhibits a strong negative association with the
probability of maintaining contacts with native Dutch. As a consequence, immigrants are less likely
to adopt Dutch attitudes and values, such as social trust. In the section ‘data and measurement’ we
construct three scales that include both a cognitive and a structural part (see below). We take these
scales as proxies for social trust in the neighborhood.
Putnam (2007, pp. 141-142) discusses two theories that deal with diversity and social
connections. According to the ‘contact hypothesis’, more diversity implies more inter-ethnic
tolerance and social solidarity. The reasoning is that ‘as we have more contact with people unlike
us, we overcome initial barriers of ignorance and hesitation and come to trust them more’. This line
of reasoning stems from the intergroup theory, which originates from Allport (1979) and is more
recently extended by Pettigrew (1998). Intergroup theory states that the contact between groups is at
a maximum when five conditions are met: equal status between groups, common goals to be
reached, intergroup cooperation, support of laws and customs and the potential to friendship.
Especially for this last condition it is necessary to take in account a longitudinal perspective. The
theory predicts prejudice to be minimal when the intergroup contact is maximal, many
investigations also found this result (see Pettigrew, 1998). Conflict-theory argues exactly the
opposite: diversity fosters out-group distrust and in-group solidarity. In other words, the more we
are brought into physical proximity to people of a different ethnicity, the less we trust the ‘other’.
Putnam introduces ‘constrict theory’, suggesting that ethnic diversity might reduce both in-
group and out-group trust. His (implicit) argumentation is that when the social context is more
diverse in terms of ethnic groups, there are more people ‘unlike you’. As a result, there are less
people with whom one can identify, resulting in fewer social connections and lower levels of trust.
That is, ethnic diversity may correlate negatively with social trust (see also Allen & Cars, 2001;
Alesina & La Ferrara, 2002). Putnam shows that for the US this is indeed the case: in
neighborhoods where ethnic diversity is higher, trust is lower.
We replicate Putnam’s analysis and test this relation for a West-European country, the
Netherlands. One might argue that the European context, and thus the mechanism between diversity
and trust, is different because European welfare states are in various ways more successful in
reducing spatial and socio-economic inequality of their citizens and probably also of their
immigrants (Esping-Anderson & Gallie, 2002; Parsons & Smeeding, 2006). For example, Musterd
(2003) analyzes the relationship between segregation and several forms of integration in
Amsterdam. It is suggested that in more polarized societies such as the US this relationship may be
stronger, whereas in Western Europe, where more moderate levels of social and spatial inequality
are predominant, this relation may be absent. Musterd finds that if there exists such a relationship in
the Netherlands, it is not a very strong one. He concludes that programs that are aimed at de-
segregation by mixing neighborhoods socially and ethnically will probably contribute little to
enhancing integration in society. Naturally, the question in this paper is whether (the absence of)
this relationship also holds for diversity and social trust.
According to Putnam (2007, pp. 138-139), the negative relation between diversity and trust is
a short-term process. In the long run, he argues, ‘societies have overcome such fragmentation by
Page 5
4
creating new compassing identities that dampen the negative effects of diversity’. He however does
not prove that claim systematically, but only by anecdotic evidence of recent successes of
overcoming ethnic fragmentation in the USA. As opposed to the European countries, the US is a
rather ‘old’ immigration country. That is, in the US, ethnic diversity has been present for a long
time. Whereas the US has dealt with flows of immigrants for a number of years, the immigration
wave of the ‘guest workers’ in Europe has started in the 1950s only (Parsons & Smeeding, 2006).
That is, if the relation as found by Putnam is indeed one of the short term, we certainly expect to
find it in the European context. In order to understand the relation between ethnic diversity and trust
better, we distinguish between two forms of trust: individual trust in neighbors and neighborhoods
and trust between other ethnic groups.
Figure 1 gives the conceptual model of ethnic diversity in the neighborhood and individual social
trust, but with our addition of the ethnic diversity of the neighbors of ego as intervening mechanism
between the diversity of the neighborhood (which is in the Netherlands around 6,000 persons large)
and the individual attitudes. The respective hypotheses are indicated in figure 1.
Figure 1 about here
We follow Putnam’s study, and we use our more elaborate conceptual model to reformulate his
general hypothesis as follows:
H1 ‘Ethnic diversity in the neighborhood negatively affects individual social trust, both for
immigrant and native residents, independently of the ethnic diversity of ego’s neighbors.’
One could argue that, besides ethnic diversity in the neighborhood affecting social trust, trust is also
affected by a more micro-level process. That is, whereas ethnic diversity in the neighborhood
explains social trust; having ethnically different neighbors explains trust as well. This is even more
important when one aims to explain trust in neighbors, as Putnam does in his main analysis. It is
likely that if there is an effect of diversity in the neighborhood, this effect will be even more
prevalent with respect to the ethnic diversity of the direct neighbors. Stolle, Soroka, & Johnston
(2008) show with Canadian and US data that not everybody in ethnically divers neighborhoods is
equally sensitive to their environment. Individuals who regularly talk with their neighbors are less
influenced by the racial and ethnic character of their surroundings than people who lack such
interaction. Therefore, we include the ethnic diversity of the neighbors, a measure of ethnic
diversity in the direct environment of ego. We assume that the mechanism is similar to that on the
neighborhood level and therefore hypothesize that:
H2 ‘Having neighbors that are ethnically different negatively affects individual social trust, both
for immigrant and native residents, independently of the ethnic diversity of the
neighborhood.’
Having direct neighbors that are of different ethnicity is likely to be influenced by the diversity in
the neighborhood itself (Bodygendrot & Martiniello, 2000). To further unravel the effect of
diversity in the neighborhood on social trust, we also examine to what extent diversity in the
neighborhood affects the likelihood of having direct neighbors that are ethnically different. This is
formulated in hypothesis three.
H3 ‘Ethnic diversity in the neighborhood increases the likelihood of having neighbors that are
ethnically different.’
If both the second and third hypothesis are upheld by our data, ethnic diversity of neighborhoods
Page 6
5
can have an indirect negative effect on individual trust, even if the first hypothesis (the direct effect)
has to be rejected. Putnam did not unravel these direct and indirect effects of ethnic neighborhoods’
diversity on individual trust, an omission that undermines the validity of his results.
3. Data and measurement
For measurement, we use the ‘Sociale Positie en Voorzieningengebruik van Allochtonen’ (Social
Position and Facilities Use of Ethnic Minorities, SPVA; (Martens, 1999). The SPVA survey is the
main data source for monitoring the disadvantage of ethnic minorities in the Netherlands
(Guiraudon, Phalet, & Ter Wal, 2005). The SPVA data provide samples of households from the four
largest immigrant ethnic minority groups in the Netherlands (Turks, Moroccans, Surinamese and
Antilleans) and a comparable native Dutch sample. The survey is a stratified sample, in which the
respondents are selected in thirteen communities with relatively large numbers of these four
minority groups. The sample aims to be representative for the Netherlands. SPVA surveys have
been conducted in 1988, 1991, 1994, 1998 and 2002. We make use of the 1998 wave because it
contains suitable indicators of trust in neighbors and the neighborhood.
Within the SPVA, an individual is classified into a minority group if he or she was born in
the respective country or if one of their parents was born there. First generation immigrants are
defined as those who are born in Turkey, Morocco, Suriname or the Dutch Antilles. Second
generation immigrants are those that are born in the Netherlands with at least one parent born in one
of the aforementioned countries, or those that are born abroad and migrated to the Netherlands at an
age younger than six.
Putnam’s neighborhood level is the USA census tract. The SPVA 1998 contains the four-
digit postal code of all respondents, which are comparable to the USA census tracts used by Putnam.
Matching the respondents to the neighborhood they live in based on their Dutch four-digit zip code
allows us to replicate Putnam’s analysis. USA Census tracts contain between 2,500 and 8,000
inhabitants and, when first delineated, are designed to be homogeneous with respect to population
characteristics, economic status, and living conditions. The average number of people that lives in a
Dutch four-digit zip code area is with 6,543 (for 1998) highly comparable to that of a USA census
tract.
Dutch four-digit zip codes overlap to a large extent with local neighborhoods (‘buurten’), as
defined by the municipalities. Neighborhoods are mostly defined based on natural borders such as
parks, rail- and waterways or major roads. Since the borders often mark building styles and -periods
as well, neighborhoods are relatively homogenous with respect to socio-demographic characteristics
(Wittebrood & Van Dijk, 2007). Since the SPVA provides the four-digit zip codes of the
respondents, we use this as the neighborhood level in our analyses2. Dutch zip code areas are
somewhat less homogenous than neighborhoods, because, as opposed to neighborhoods, the borders
of Dutch zip code areas are defined to facilitate the postal distribution process, not as an indication
of local neighborhoods. Furthermore, the size of the population in a zip code areas is very similar
across the Netherlands (Wittebrood & Van Dijk, 2007), while the size of the population s of
neighborhoods varies more, just like the USA census tracts. The Dutch Statistical Office provides a
number of relevant indicators on the zip code level (see below).
The dependent variables
For the US, Putnam (2007) uses variables such as inter-ethnic trust, trust in neighbors, intra-ethnic
trust, confidence in local government, less expectation that others will cooperate, working in
community project, giving to charity or volunteering, having close friends and confidents, less
happiness and quality of life, more time spending on TV watching.
We construct three similar measures for the measurement of social trust. First, we construct a
scale that measures trust in one’s direct neighbors. Second, we build a scale that measures trust in
Page 7
6
one’s neighborhood. These scales contain items on the frequency, opinion and the quality of contact
with one’s direct neighbors and people in the neighborhood. The third dependent variable can be
seen as measure of social distance or inter-ethnic trust, containing two items measuring the
respondent’s opinion on the background of the friends and the partner of one’s children, ranging
from very disturbing to not disturbing at all (comparable to some items of the classical social
distance scale, developed by Bogardus, (1925, 1928, 1933). Each of the scales has the same
psychometric characteristics for the four immigrant groups and for the native Dutch. A detailed
description of the construction of the scales and their psychometric characteristics can be found in
the appendix. These three indicators are referring to two different forms of trust: the first two refer
to individual trust in neighbors and neighborhoods and the third to trust between other ethnic
groups. The correlations between the three indicators underline this difference: between the first two
it is .50, between the first two and the last zero (see table A7).
Independent variables at the individual level
The controls included on the individual level are virtually equal to those used by Putnam. We
control for the following: gender, age, marital status, educational attainment, family income and
satisfaction with current finances3, home ownership, citizenship, ethnicity and being a second-
generation immigrant. Furthermore, to control for the language proficiency of the respondent we
constructed a Mokken scale4 (see the appendix, table A8 and A9 for the construction of this scale).
As mentioned in section two, the SPVA survey also provides information on the direct
neighbors of the respondent. As an extra measure of ethnic diversity on the sub-neighborhood level,
we include the ethnicity of the direct neighbor(s). This is operationalized as whether or not (all) the
neighbor(s) of the respondent belong to the same ethnic group as the respondent.
Independent variables at the zip code and municipality level
On the zip code level we include the following controls, obtained from the ‘Kerncijfers wijken en
buurten 2004’, provided by the Dutch Statistical Office. First, based on the percentage of ethnic
groups that lives in the respective zip code area, we construct a Herfindahl index of ethnic diversity
(range: -1 till 0)5. The index is interpreted as follows: a value of -1 implies no diversity at all, i.e. the
neighborhood consists of one ethnic group only. A value approaching zero means total diversity: all
people in the neighborhood have a different ethnicity. This index is read as the inverse of the index
of ethnic homogeneity as used by Putnam.
Second, we include the mean income, an income inequality ratio6, the population density, the
number of people moving house from or to the neighborhood7, the percentage of renters, the
percentage of students in higher education, and the percentage of people over 65 years old. On the
municipality level, we include the Gini-coefficient for income8, and the percentage of non-violent
and violent crimes9. These neighborhood characteristics are more or less comparable with those
used by Putnam.
4. Results
In table 1A, a descriptive overview of the individuals of the sample is presented. The averages of
the three indicators of trust are not very different for natives and immigrants. But the averages of
age, income and satisfaction with current finances of the natives are higher than those of the
immigrants. The educational levels of natives and immigrants deviate strongly, the natives being
higher educated than the immigrants. Immigrants are more often male and married. House
ownership is more frequent among natives and they have less often neighbors of another ethnic
group. The bottom part of table 1A gives the distribution between the five ethnic groups in the
sample.10
INSERT TABLE 1A & 1B
Page 8
7
Table 1B give the descriptive statistics of the variables at the zip code and municipality levels. In
figure two, three and four, the mean Herfindahl index of ethnic diversity in a municipality is plotted
against each of our dependent variables. These figures are similar to those presented by Putnam
(2007): in the municipalities where ethnic diversity is higher, trust is lower. The case of inter-ethnic
trust shows a slight positive relationship at the municipality level.
Figures 2 to 4 about here
Naturally, this is simply a bivariate representation of the relationship in question, and these results
might be spurious because differences in the individual and neighborhood characteristics might be
responsible for these bivariate relations.
In the following tables we present multi-level regression analyses including the relevant controls at
the individual and the neighborhood level.
Trust in neighbors
Table two presents a multilevel regression model predicting trust in direct neighbors. Model one
only contains the Herfindahl Index of Ethnic Diversity; its coefficient is negative and significant. In
model two, the ethnic groups and a dummy for second-generation immigrants are added. Whereas
the Turks and the Moroccans do not trust their neighbors differently from the native Dutch, the
Surinamese, Antilleans and the second-generation immigrants trust their neighbors less than the
Dutch do. Controlling for the remaining individual characteristics (model three), we see that the
effect of ethnic diversity hardly diminishes and remains significantly negative. Furthermore, Turks
and Moroccans trust their neighbors significantly more than the native Dutch, while second
generation immigrants still have less trust in their neighbors as the native Dutch. There is no effect
of educational attainment on trust, besides that of those who didn’t provide information on their
educational degree. Last, we see that those that are older, married or own a house have significantly
more confidence in their neighbors. Model four and five respectively introduce the neighborhood
and municipality characteristics, just like Putnam. None of these controls appears significant and
their inclusion does not change the effect of neighborhoods’ ethnic diversity on thrust in neighbors.
In model six the ethnicity of the neighbor is added. As expected, this significantly and negatively
predicts trust in one’s neighbors: if the ethnicity of the neighbors is different of respondents’ own,
the trust in neighbors is lower. However, the effect of neighborhoods’ ethnic diversity remains
unchanged by this inclusion. This results supports both our first and second hypotheses: living in an
more ethnical devours neighborhood and having neighbors of another ethnic group both decreases
independently of each other the trust in one’s neighbors. Last, in model seven, an interaction term
between the Herfindahl Index and the ethnicity of the neighbors is included: this interaction term is
significant, while the effect of the Herfindahl Index becomes insignificant and that of neighbors’
ethnicity even increases. This means that having ethnically different neighbors always decreases
trust in one’s neighbors, and that this decline of trust is steeper if ones has ethnically different
neighbors in more ethnically diverse neighborhoods. The inclusion of this interaction term does not
change the support of our second hypothesis, but qualifies the support of the first hypothesis. The
later is only true if one lives next to neighbors coming from another ethnic group than ones own.
We also tested whether the relation between ethnic diversity in the neighborhood and individual
trust is different for the different ethnic groups by including the relevant interaction terms. We
found no significant interaction terms (not shown here). This means that the results are true for both
natives and immigrants alike, just as Putnam found: ethnic diversity in neighborhoods decreases
trout in neighbors for immigrants and natives alike.
Table 2 about here
Page 9
8
Trust in neighborhood
In table three we present an analogous multilevel analysis explaining trust in the neighborhood.
Model one shows a negative uni-variate relationship between ethnic diversity and trust in the
neighborhood. In model two it appears that Turks trust their neighborhood more than native Dutch,
Antilleans and second-generation immigrants less. When including the individual level
characteristics (model three), only Turks score significantly higher on trust. Ethnic diversity in the
neighborhood significantly predicts a lower trust in the neighborhood. Furthermore, upper
secondary and college/university graduates trust their neighborhood less than those with a primary
education only. As in the previous analysis, being married, older or owning a house positively
predicts trust in the neighborhood. In model four and five, the neighborhood and municipality
characteristics are added. Although none of the added variables but population density appear
significant, ethnic diversity is no longer significant, probably due to the correlation between these
neighborhood variables. In model six, the ethnic diversity of the neighbors is added. As expected,
this negatively predicts trust in the neighborhood. Also with respect to trust in the neighborhood it
was tested with interaction terms whether or not the relation between ethnic diversity and trust in the
neighborhood is different for the ethnic groups. This appeared not to be the case (not shown here):
the relation between ethnic diversity and trust in the neighborhood is the same for immigrant and
native residents. Neither there was a significant interaction term between the Herfindahl Index and
the ethnicity of the neighbors on trust in the neighborhood.
Model seven is a cleaned model with only significant variables: neighborhood ethnic
diversity, population density and the different ethnicity of neighbors decreases trust in
neighborhoods, while a Turkish respondent, a higher age, being married and home ownership
increase trust in neighborhoods.
Table 3 about here
Inter-ethnic trust
In table three we present an analogous multilevel analysis explaining inter-ethnic trust. Model one
shows a negative uni-variate relationship between ethnic diversity and trust in the neighborhood.
Model two shows that Surinamese and Antilleans, but also second-generation immigrants, have
higher inter-ethnic trust than natives, while second generations immigrants have higher inter-ethnic
trust as first generation or natives. After addition of individual characteristics (which all have the
usual effect on inter-ethnic trust) the negative effect of ethnic neighborhoods’ diversity becomes
insignificant. This later result is remarkable and it remains insignificant also after the addition of zip
code and municipality variables. Having ethnically different neighbors increases one’s interethnic
trust as model six shows. Also with respect to inter-ethnic trust we tested with interaction terms
whether or not the relation between ethnic diversity and inter-ethnic trust is different for the ethnic
groups. This appeared not to be the case (not shown here): the relation between ethnic diversity and
inter-ethnic trust is the same for immigrant and native residents. Neither there was a significant
interaction term between the Herfindahl Index and the ethnicity of the neighbors on inter-ethnic
trust.
Our results support that trust in neighbors and trust in neighborhood cannot be equalized with
inter-ethnic trust. The first two forms of trust are negatively influenced by the ethnic diversity of the
neighborhood and neighbors, while the latter is positively influence by having ethnically different
neighbors.
Table 4 about here
The ethnicity of the neighbors
In table five, the last link of our conceptual model is tested: to what extent does ethnic diversity in
the neighborhood predict the ethnic diversity of the neighbors? Contrary to the early models we use
Page 10
9
logistic regression, in which parameters higher than 1.00 indicate that the dependent score is higher
if the score on the independent score increases and parameters lower than 1.00 indicates that the
dependent score is lower if the score on the independent score increases.
In model one, we see that without any controls living in an ethnic diverse neighborhood, the
likelihood of having neighbors that are ethnically different increases. In model two, we see that the
odds of having ethnically different neighbors can to be fully explained by belonging to one of the
four immigrant groups. Independently of the ethnic diversity of a neighborhood, immigrants have
much higher odds to live next to another ethnic neighbor than natives. Note that this does not mean
that immigrants have a higher chance to live with a native neighbor, it only means that immigrants
are less likely to live next to neighbor of the same ethnic group. However, it also means that natives
live far more often next to natives, even in ethnically diverse neighborhoods than any immigrant
group. However, also when adding the remaining individual characteristics (model three), the ethnic
diversity in the neighborhood is a significant but negative predictor for the ethnic diversity of the
neighbors. The effect of neighborhoods’ ethnic diversity on the likelihood of having ethnically
different neighbors becomes insignificant when adding zip code and municipality characteristics,
especially due to the variable ‘moving house mobility’ (models 4 and 5). These results partly
support our third hypothesis: living in ethnically diverse neighborhoods increases of the likelihood
of having neighbors that are ethnically different, but this is only true for the natives. We also ran the
analysis containing only the natives. For them, ethnic diversity in the neighborhood (including all
controls) does predict the likelihood of having ethnically different neighbors significantly positive.
Given the strong effects of having ethnically diverse neighbors on individual trust in neighbors and
neighborhoods, the partial confirmation of our third hypothesis means that a part of the effect of
ethically diverse neighborhoods on individual trust is intermediated by the ethnic diversity of one’s
neighbors.
Table 5 about here
5. Discussion
The first conclusion of our study is that the claim of Putnam (2007) that immigration and ethnic
diversity tend to reduce solidarity and social capital can also be tested for Europe. In assuming that
the right data are not available in Europe, Giddens (2007) was wrong. The data necessary to test this
claim are available, but often out of reach of social scientists due to stricter European privacy laws
and a larger hesitance of European social scientists to ask publicly political incorrect questions.
The second more important conclusion is that Putnam’s USA conclusions about the negative
effect of ethnically diverse neighborhoods on individual trust are also valid for European welfare
states, like the Netherlands. Putnam’s results are thus not unique for immigrant societies like the
USA, Canada or Australia (Leigh, 2006). At least at the short term there exists a negative relation
between ethnic diversity of neighborhoods and neighbors on the one hand and trust in neighbor and
the neighborhoods at the other hand. This relation cannot be explained away by the characteristics
of the respondents, or characteristics of the zip code area or municipalities where they live. It will be
not easy to falsify this conclusion; neither to portrait our Dutch results as an European outlier. The
position of immigrants in the Netherlands is not exceptional in comparison with immigrants in other
European societies (Fleischmann & Dronkers, 2007; Andre, Dronkers, & Fleischmann,
Forthcoming). Hence there is no reason to assume that our conclusion is not true for other EU
member states.
However, individual trust in neighbors and neighborhoods is something else than trust in
other ethnic groups than one’s own. Individual trust in neighbors and neighborhoods does not
correlate significantly with inter-ethnic trust. This is reflected in our results: the neighborhood’s
ethnic diversity has no negative effect on the level of interethnic-trust; the same holds for the ethnic
diversity of neighbors. In contrast, having ethnically different neighbors increases inter-ethnic trust,
as predicted by the ‘Intergroup theory’. Inter-group theory originates from Allport (1979) and is
Page 11
10
lately extended by Pettigrew (1998). Intergroup theory states that the positive impact of contact
between different (ethnic) groups is at a maximum when five conditions are met: equal status
between groups, common goals to be reached, inter-group cooperation, support of laws and customs
and the potential to friendship. Contact between neighbors might meet a few of these conditions of a
positive impact of contact between different (ethnic) groups, like equal status, common goals and
cooperation.
Often European political elites find it difficult to distinguish between these different forms of
(dis)trust and lump them together as racism. As a consequence they deny or censure complains
about the negative effects of ethnic diversity on the quality of neighborhoods as manifestations of
racism. This denial or censure of the negative effects of neighborhoods’ ethnic diversity on wrong
grounds (racism) might explain the disenchantment of natives in ethnically diverse neighborhoods
of mainstream politics and their tendency for extreme voting behavior.
Another interesting conclusion from our analysis is that policies aiming at promoting
ethnically diverse neighborhoods in order to promote ethnic integration at the societal level might
have an unintended inverse effect of decreasing individual trust in these neighborhoods (compare
with Musterd, 2003). Moreover, living in ethnically diverse neighborhoods only increases the odds
of living next to a neighbor of another ethnic group only for the natives, not for immigrant residents.
Natives in ethnically diverse neighborhoods have a greater likelihood of have non-native neighbors,
but also immigrants in ethnically diverse neighborhoods have a lower chance of living next to
neighbors from another ethnic group , there is no effect.
The explanation of these results might be the high importance of bridging social capital,
especially in ethnic diverse neighborhoods in combination with the higher cost of forming bridging
social capital. The lack of bridging social capital in ethnically divers neighborhoods seems to be a
more important explanation than the amount of bonding social capital. A related explanation might
be the language diversity in ethnically diverse neighborhoods, which makes the forming of bridging
social capital difficult and thus decreases trust (Leigh, 2006). In a next paper we will try to test this
explanation.
Page 12
11
Notes
1 Leigh (2006) made a comparable analysis as Putnam did, using Australian data. He found similar results as Putnam,
but he found that the effect of linguistically diverse neighborhoods was for stronger than that of ethnically diverse
neighborhoods. We will test this explanation in a sequel of this paper.
2 Some contextual information is only available on the neighborhood level and not on the zip code level. This is dealt
with in the following way. The Dutch Bureau of Statistics provides the most frequent zip code for all neighborhoods.
Hence, respondents are matched to the most frequent zip code. However, since neighborhoods can be small, some have
the same most frequent zip code. When this is the case, neighborhoods are aggregated to the zip code level. This is done
by averaging the indicators on the neighborhood level, weighted by the number of people that live in the respective
neighborhood. Those zip code areas which contain only one SPVA respondent were excluded from the analyses (20
cases).
3 Coded on a four-point scale; the exact wording of the item is: ‘What do you think of your income/your family income,
is it: more than enough, enough, too little, or by far too little to live off?’
4 The language items were not included for the native Dutch respondents, since it is their mother tongue,. Therefore, the
Dutch natives were given the highest score on the scale.
5 Putnam calculates an index of Ethnic Homogeneity. Since the topic at hand is diversity rather than homogeneity, we
choose to use an index of ethnic diversity, which is the Herfindahl index as used by Putnam, multiplied by -1. The
Herfindahl index of ethnic diversity is calculated as follows: -((fraction ethnic group 1)2 + (fraction of ethnic group 2)2
+….+ (fraction of ethnic group n)2). The following ethnic groups are included: Turks, Moroccans, Antilleans,
Surinamese, Other non-western immigrants, Western immigrants, native Dutch. The data used is the ‘Kerncijfers wijken
en buurten 2004’, obtained from the Dutch Bureau of Statistics (www.cbs.nl).
6 The calculated neighborhood income inequality ratio is the percentage of people with an income lower than or equal to
the 40 percentage points of the national income distribution, divided by the percentage of people with an income above
or equal to the 80 percentage points of the national income distribution.
7 The number of people moving house is expressed as: the number op people per 1000 inhabitants of the neighborhood,
who have moved from or to the neighborhood in the past year. This number consists of the number of intra-
neighborhood moves, plus half of the sum of the leavers, plus half of the sum of the arrivals in the respective
neighborhood.
8 The Gini-coefficient is not available on the neighborhood level and therefore included on the municipality level.
9 The percentage of violent and non-violent crimes is expressed as a fraction of the total population in the municipality.
10 Only 512 of the 4614 immigrants are second-generation immigrants.
Page 13
References
Aguilera, M. B. (2002). The impact of social capital on labor force participation: evidence from the
2000 social capital benchmark survey. Social Science Quarterly, 83(3), 854-874.
Alesina, A., & La Ferrara, E. (2000). Participation in heterogeneous communities. The Quarterly
Journal of Economics, 115(3), 847-858.
Alesina, A., & La Ferrara, E. (2002). Who trusts others? Journal of Public Economics, 85(2), 207-
234.
Allen, J., & Cars, G. (2001). Multiculturalism and governing neighbourhoods. Urban Studies,
38(12), 2195-2209.
Allport, G. W. (1979). The nature of prejudice. Cambridge, MA: Perseus Books.
Andre, S., Dronkers, J., & Fleischmann, F. (Forthcoming). The Different Levels of Discrimination,
experienced by First and Second Generation Immigrants from Different Countries of Origin
in the different EU member-states. Florence: EUI.
Bodygendrot, S., & Martiniello, M. (2000). Minorities in European cities: the dynamics of social
integration and social exclusion at the neighbourhood level. London: Macmilan.
Bogardus, E. S. (1925). Measuring social distance. Journal of Applied Sociology, 9, 299-308
Bogardus, E. S. (1928). Immigration and Race Attitudes. Washington DC: Heath
Bogardus, E. S. (1933). A social distance scale. Sociological and Social Research 17(1), 265-271.
Bourgeois, M., & Friedkin, N. E. (2001). The distant core: social solidarity, social distance and
interpersonal ties in core-periphery structures. Social Networks, 23(4), 245-260.
Brisson, D. S., & Usher, C. L. (2007). The effects of informal neighborhood bonding social capital
and neighborhood context on homeownership for families living in poverty. Journal of
Urban Affairs, 29(1), 65-75.
Burt, R. S. (2001). Structural holes versus network closure as social capital. In N. Lin, K. Cook &
R. S. Burt (Eds.), Social Capital: Theory and Research (pp. 31-56). New York: Aldine De
Gruyter.
Burt, R. S. (2002). Bridge decay. Social Networks, 24(4), 333-363.
Dagevos, J., & Ode, A. (2003). Minderheden in Amsterdam. Contacten, concentratie en integratie.
Amsterdam: SISWO.
Dekker, K. (2007). social capital, neighbourhood attachment and participation in distressed urban
areas. a case study in the Hague and Utrecht, the Netherlands. Housing Studies, 22(3), 355-
379.
Esping-Anderson, G., & Gallie, D. (Eds.). (2002). Why we need a welfare state. Oxford: OUP.
Fennema, M., & Tillie, J. (1999). Political participation and political trust in Amsterdam: civic
communities and ethnic networks. Journal of Ethnic and Migration Studies, 25(4), 703-726.
Fleischmann, F., & Dronkers, J. (2007). The effects of social and labour market policies of EU-
countries on the socio-economic integration of first and second generation immigrants form
different countries of origin. (EUI/RSCAS Working Paper, No. 19). Florence: EUI.
Gesthuizen, M., Van der Meer, T., & Scheepers, P. (Forthcoming). Ethnic diversity and social
capital in Europe: tests of Putnam's thesis in European countries. Scandinavian Political
Studies.
Giddens, A. (2007). Doubting diversity’s value. Foreign Policy, (November/December), 86-88.
Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology, 78(6), 1360-
1380.
Granovetter, M. (1995). Getting a job. A study of contacts and careers (2nd ed.). Chicago:
University of Chicago Press.
Guiraudon, V., Phalet, K., & Ter Wal, J. (2005). Monitoring ethnic minorities in the Netherlands.
International Social Science Journal, 57(1), 75-88.
Hooghe, M., Reeskens, T., Stolle, D., & Trappers, A. (2006). Ethnic diversity, trust and
ethnocentrism and Europe. A multilevel analysis of 21 countries. Paper presented at the
Page 14
13
'Annual Meeting of the Political Science Association', Philadelphia, August 31st-September
3rd 2006.
Lancee, B. (Forthcoming). The economic returns of immigrants' bonding and bridging social
capital. A case study in the Netherlands, Paper to be presented at the 103rd Annual Meeting
American Sociological Association, Boston, August 1-4th 2008.
Leigh, A. (2006). Trust, inequality and ethnic heterogeneity. The Economic Record, 82(258), 268-
280.
Lin, N. (1999). Social networks and status attainment. Annual Review of Sociology, 25, 467-487.
Martens, E. P. (1999). Minderheden in beeld, SPVA-98. Rotterdam: Instituut voor Sociologisch-
Econmisch Onderzoek (ISEO).
Mokken, R. J. (1996). Nonparametric models for dichotomous responses. In W. J. Van Der Linden
& R. K. Hambleton (Eds.), Handbook of Modern Item Response Theory (pp. 351-367). New
York: Springer.
Molenaar, I. W., & Sijtsma, K. (2000). User’s manual MSP5 for windows: a program for Mokken
scale analysis for polytomous items. Groningen: ProGAMMA.
Murie, A., & Musterd, S. (2004). Social exclusion and opportunity structures in European cities and
neighbourhoods. Urban Studies, 41(8), 1441-1459.
Musterd, S. (2003). Segregation and integration: a contested relationship. Journal of Ethnic and
Migration Studies, 29(4), 623-641.
Parsons, C. A., & Smeeding, T. M. (2006). Immigration and the transformation of Europe.
Cambridge: Cambridge University Press.
Pettigrew, T. (1998). Intergroup contact theory. Annual Review of Psychology, 49, 65-85.
Poortinga, W. (2006). Social relations or social capital? Individual and community health effects of
bonding social capital. Social Science and Medicine, 63, 255-270.
Portes, A. (2000). The two meanings of social capital. Sociological Forums, 15(1), 1-12.
Putnam, R. D. (2007). E Pluribus Unum: Diversity and community in the twenty-first century. The
2006 Johan Skytte prize lecture. Scandinavian Political Studies, 30(2), 137-174.
Sijtsma, K., & Van der Ark, L. A. (2003). Investigation and treatment of missing item scores in test
and questionnaire data. Multivariate Behavioral Research, 38, 505-528.
Snijders, T. A. B., & Bosker. (1999). Multilevel analysis. An introduction to basic and advanced
multilevel modeling. London: Sage Publications.
Stolle, D., Soraka, S., & Johnston, R. (2008). When does diversity erode trust? Neighborhood
diversity, interpersonal trust and the mediating effect of social interactions. Political Studies,
56, 57-75.
Van der Laan Bouma-Doff, W. (2007). Confined contact: residential segregation and ethnic bridges
in the Netherlands. Urban Studies, 44(5), 997-1017.
Van Ginkel, J. R., & Van der Ark, L. A. (2007). SPSS syntax for two-way imputation of missing
test data. Retrieved 13-9, 2007, from
http://www.uvt.nl/faculteiten/fsw/organisatie/departementen/mto/software2.html
Van Schuur, W. H. (2003). Mokken scale analysis: between the Guttman scale and parametric item
response theory. Political Analysis, 11, 139-163.
Van Schuur, W. H., & Kiers, H. A. L. (2004). Why factor analysis often is the incorrect model for
analyzing bipolar concepts and what model to use instead. Applied Psychological
Measurement, 18(2), 97-110.
Wildegren, O. (1997). Social solidarity and social exchange. Sociology, 31(4), 755-771.
Wittebrood, K., & Van Dijk, T. (2007). Aandacht voor de wijk. Effecten van herstructurering op de
leefbaarheid en de veiligheid. The Hague: Sociaal Cultureel Planbureau.
Page 15
14
Figure 1. A conceptual model of ethnic diversity and individual social trust.
Figure 2. Ethnic diversity and trust in neighbors.
Ethnic diversity
in the
neighborhood
Ethnicity diversity
of direct
neighbors
Individual social
trust
+
-
-
Hypothesis 1
Hypothesis 3
Hypothesis 2
Page 16
15
Figure 3. Ethnic diversity and trust in the neighborhood.
Figure 4. Ethnic diversity and inter-ethnic trust.
Page 17
16
Table 1A. Descriptive statistics individual level variables.
Native Dutch
Immigrants
Immigrants+
Natives
Mean
S.d.
Mean
S.d.
Mean
S.d.
Trust in direct neighbors
0.63
0.13
0.61
0.14
0.61
0.14
Trust in neighborhoods
0.65
0.18
0.62
0.18
0.63
0.18
Inter-ethnic trust
0.7
0.22
0.73
0.27
0.72
0.26
Age
48.73
17.66
39.05
12.76
40.97
14.39
Language proficiency
1
0
0.61
0.33
0.69
0.34
Family income
3404.75 2071.83 2653.95 1477.73 2803.12 1640.63
Satisfaction with current
finances
0.6
0.24
0.42
0.25
0.45
0.26
%
N
%
N
%
N
Married
38.06
435
43.54
2009
42.55
2444
Female
47.94
548
40.83
1884
42.24
2432
Educational Degree
Primary
22.57
258
46.53
2147
41.78
2,405
Lower Secondary
24.15
276
19.9
918
20.74
1,194
Upper Secondary
24.41
279
20.65
953
21.4
1,232
College/University
26.33
301
9.97
460
13.22
761
No info on educational level
2.54
29
2.95
136
2.87
165
Dutch citizen
100
1,143
71.69
3,308
77.31
4,451
Owns house
41.03
469
13.26
612
18.78
1,081
Ethnicity neighbor(s) different
4.9
56
77.2
3,562
62.85
3,618
Distribution between ethic
groups
%
N
Turks
18.78
1,081
Moroccans
19.42
1,118
Surinamese
26.06
1,500
Antilleans
15.89
915
Native Dutch
19.85
1,143
Sample
100
5,757
Source: SPVA 1998
Page 18
17
Table 1B. Descriptive statistics zip code and municipality level variables.
Mean
Sd Range
Zip code level Herfindahl Index
-0.29
0.20
-1-0
(N=291)
% In higher
education
0.19
0.14
0-1
Mean income per
person
0.48
0.09
0-1
Inequality ratio
0.31
0.18
0-1
Population density
0.37
0.2
0-1
% Over 65 years old
0.33
0.15
0-1
% Renters
0.69
0.2
0-1
Moving house
mobility
0.44
0.12
0-1
Municipality
level
Gini coefficient
0.53
0.24
0-1
(N=13)
% Violent crimes
0.58
0.37
0-1
% Non-violent
crimes
0.59
0.31
0-1
Source: CBS Statline
Page 19
18
Table 2. Multilevel linear regression predicting trust in direct neighbors, Nindiv=5,757; Nzipcode=
291; Nmun= 13 (standardized coefficients, standard errors between brackets).
Model 1
Model
2
Model
3
Model
4
Model
5
Model
6
Model
7
Herfindahl Index of Ethnic Diversity -.092***
-
.093***
-
.070***
-.058** -.057** -.060** -.035
(.011)
(.011) (.011) (.021) (.022) (.022) (.025)
Dutch
ref.
ref.
ref.
ref.
ref.
ref.
Turkish
.005 .045*** .044*** .044*** .079*** .073***
(.006) (.009) (.009) (.009) (.010) (.010)
Moroccan
-.006 .031*** .031*** .031*** .069*** .063***
(.006) (.009) (.009) (.009) (.009) (.010)
Surinamese
-.012* .008 .007 .007 .044*** .038***
(.006) (.006) (.006) (.006) (.007) (.008)
Antillean
-
.028***
.004 .003 .003 .047*** .041***
(.006) (.007) (.007) (.007) (.008) (.009)
Second generation
-
.025***
-.018* -.017* -.017* -.017* -.018*
(.007) (.007) (.007) (.007) (.007) (.007)
Female
.001 .001 .001 -.001 -.001
(.004) (.004) (.004) (.004) (.004)
Age
.089*** .091*** .091*** .092*** .092***
(.011) (.012) (.012) (.011) (.011)
Married
.023*** .022*** .023*** .021*** .021***
(.005) (.005) (.005) (.005) (.005)
Primary education
ref.
ref.
ref.
ref.
ref.
Lower secondary
.010 .010 .010 .009 .009
(.005) (.005) (.005) (.005) (.005)
Upper secondary
.009 .009 .009 .010 .010
(.006) (.006) (.006) (.006) (.006)
College/University
.004 .005 .005 .007 .006
(.007) (.007) (.007) (.007) (.007)
No info on educational level
.023* .022* .022* .024* .024*
(.011) (.011) (.011) (.011) (.011)
Language proficiency
.037*** .037*** .037*** .039*** .038***
(.009) (.009) (.009) (.009) (.009)
Dutch citizen
.005 .005 .005 .006 .005
(.006) (.006) (.006) (.006) (.006)
Family income
.019 .019 .020 .018 .018
(.021) (.022) (.022) (.021) (.021)
Dummy imputation family income
-.002 -.002 -.002 -.002 -.002
(.005) (.005) (.005) (.005) (.005)
Home ownership
.023*** .023*** .024*** .024*** .024***
(.005) (.006) (.006) (.006) (.006)
Satisfaction with current finances
.009 .010 .010 .011 .011
(.008) (.008) (.008) (.008) (.008)
% in higher education
-.000 .007 .012 .011
(.020) (.020) (.020) (.020)
Mean income per person
-.061 -.059 -.049 -.050
(.037) (.037) (.037) (.037)
Page 20
19
Inequality ratio
-.025 -.022 -.021 -.021
(.017) (.017) (.017) (.017)
Population density
-.027 -.024 -.026 -.025
(.014) (.015) (.015) (.015)
% over 65 years old
-.004 -.002 .004 .005
(.018) (.018) (.018) (.018)
% renters
-.001 -.006 -.009 -.008
(.017) (.017) (.017) (.017)
Moving house mobility
.005 .007 -.001 -.001
(.025) (.025) (.025) (.025)
Gini coefficient
-.021 -.019 -.019
(.012) (.012) (.012)
% violent crimes
.024 .018 .018
(.026) (.026) (.026)
% non-violent crimes
-.027 -.023 -.023
(.032) (.031) (.031)
Ethnicity neighbor(s) different
-
.047***
-
.056***
(.005) (.007)
Herfindahl*Ethnicity neighbors
different
-.041*
(.020)
Constant
.587***
.597*** .493*** .543*** .555*** .550*** .559***
(.004)
(.006) (.014) (.031) (.032) (.031) (.032)
Log-likelihood
3192.3
3219.3 3297.5 3301.7 3303.3 3350.3 3352.2
Source: SPVA 1998, Dutch Bureau of Statistics
* p<0.05, ** p<0.01, *** p<0.001
Page 21
20
Table 3. Multilevel linear regression predicting trust in the neighborhood, Nindiv=5,757;
Nzipcode= 291; Nmun= 13 (standardized coefficients, standard errors between brackets).
Model 1 Model
2
Model
3
Model 4 Model
5
Model 6 Model 7
Herfindahl Index of Ethnic
Diversity
-
.120***
-
.123***
-
.093***
-.039
-.037 -.039 -
.065***
(.016) (.016) (.016) (.029) (.030) (.031) (.018)
Dutch
ref.
ref.
ref.
ref.
ref.
ref.
Turkish
.024** .043*** .043*** .042*** .062*** .034***
(.008) (.012) (.012) (.012) (.012) (.006)
Moroccan
-.013 .004
.004
.004
.026*
(.008) (.011) (.011) (.011) (.012)
Surinamese
-.014 .007
.006
.006
.028**
(.007) (.008) (.008) (.008) (.009)
Antillean
-
.032***
.005
.005
.005
.030**
(.008) (.009) (.009) (.009) (.011)
Second generation
-.019* .003
.004
.004
.004
(.008) (.009) (.009) (.009) (.009)
Female
.005
.004
.004
.003
(.005) (.005) (.005) (.005)
Age
.102*** .101*** .101*** .102*** .099***
(.015) (.015) (.015) (.015) (.013)
Married
.038*** .037*** .037*** .036*** .033***
(.006) (.006) (.006) (.006) (.005)
Primary education
ref.
ref.
ref.
ref.
Lower secondary
-.010 -.011
-.011 -.011
(.007) (.007) (.007) (.007)
Upper secondary
-.015* -.015* -.015* -.014*
(.007) (.007) (.007) (.007)
College/University
-.022* -.021* -.021* -.020*
(.009) (.009) (.009) (.009)
No info on educational
level
.018
.018
.018
.019
(.014) (.014) (.014) (.014)
Language proficiency
.030** .028* .028* .030*
(.012) (.012) (.012) (.012)
Dutch citizen
-.009 -.009
-.009 -.009
(.007) (.007) (.007) (.007)
Family income
.002
.000
.001
.001
(.027) (.028) (.028) (.028)
Dummy missing family
income
.006
.005
.006
.006
(.007) (.007) (.007) (.007)
Home ownership
.046*** .047*** .047*** .047*** .045***
(.007) (.007) (.007) (.007) (.006)
Satisfaction current
finances
.009
.009
.009
.010
(.010) (.010) (.010) (.010)
% in higher education
.010
.019
.022
(.027) (.028) (.028)
Page 22
21
Mean income per person
-.046
-.043 -.038
(.050) (.050) (.051)
Inequality ratio
-.038
-.038 -.037
(.025) (.025) (.025)
Population density
-.037
-.043* -.044* -.053**
(.020) (.021) (.021) (.018)
% over 65 years old
.028
.022
.025
(.025) (.025) (.025)
% renters
-.021
-.022 -.023
(.023) (.023) (.023)
Moving house mobility
-.020
-.025 -.030
(.036) (.036) (.036)
Gini coefficient
-.014 -.013
(.017) (.017)
% violent crimes
.064
.061
(.038) (.038)
% non-violent crimes
-.069 -.067
(.045) (.046)
Ethnicity neighbor(s)
different
-
.027***
-
.017***
(.006) (.005)
Constant
.592*** .600*** .518*** .595*** .611*** .608*** .578***
(.006) (.008) (.018) (.042) (.044) (.044) (.013)
Log-likelihood
1731.8 1761.5 1858.6 1864.2 1865.8 1875.2 1858.2
Source: SPVA 1998, Dutch Bureau of
Statistics
* p<0.05, ** p<0.01, ***
p<0.001
Page 23
22
Table 4: Multilevel linear regression predicting inter-ethnic trust, Nindiv=5,757; Nzipcode= 291;
Nmun= 13 (standardized coefficients, standard errors between brackets).
Model
1
Model
2
Model
3
Model
4
Model
5
Model
6
Herfindahl Index of Ethnic
Diversity
-
.116***
-.053* -.031 .014 .036 .038
(.029) (.021) (.020) (.038) (.039) (.039)
Dutch
ref.
ref.
ref.
ref.
ref.
Turkish
-
.121***
.024 .026 .025 -.001
(.010) (.014) (.014) (.014) (.015)
Moroccan
-
.109***
.031* .032* .032* .004
(.010) (.014) (.014) (.014) (.015)
Surinamese
.156*** .187*** .188*** .188*** .161***
(.009) (.010) (.010) (.010) (.011)
Antillean
.191*** .223*** .223*** .223*** .191***
(.010) (.011) (.011) (.011) (.013)
Second generation
.049*** -.009 -.009 -.009 -.009
(.010) (.011) (.011) (.011) (.011)
Female
.011 .010 .011 .012
(.007) (.007) (.007) (.007)
Age
-.025 -.028 -.028 -.029
(.018) (.018) (.018) (.018)
Married
-
.034***
-
.033***
-
.033***
-
.032***
(.007) (.007) (.007) (.007)
Primary education
ref.
ref.
ref.
ref.
Lower secondary
.021* .021* .020* .021*
(.008) (.008) (.008) (.008)
Upper secondary
.040*** .039*** .039*** .038***
(.009) (.009) (.009) (.009)
College/University
.072*** .068*** .068*** .067***
(.010) (.011) (.011) (.011)
No info on educational level
.009 .008 .009 .008
(.017) (.017) (.017) (.017)
Language proficiency
.135*** .134*** .135*** .133***
(.014) (.014) (.014) (.014)
Dutch citizen
.039*** .040*** .040*** .040***
(.009) (.009) (.009) (.009)
Family income
.018 .012 .012 .013
(.033) (.034) (.034) (.034)
Dummy imputation family
income
-
.046***
-
.047***
-
.047***
-
.046***
(.008) (.008) (.008) (.008)
Home ownership
.011 .012 .012 .012
(.009) (.009) (.009) (.009)
Satisfaction with current
finances
.004 .003 .003 .002
(.012) (.012) (.012) (.012)
% in higher education
.024 .015 .012
Page 24
23
(.036) (.037) (.037)
Mean income per person
.097 .115 .108
(.064) (.065) (.064)
Inequality ratio
-.041 -.037 -.038
(.032) (.032) (.032)
Population density
-.012 -.009 -.007
(.027) (.028) (.028)
% over 65 years old
-.029 -.027 -.030
(.032) (.033) (.033)
% renters
.011 .006 .008
(.030) (.030) (.030)
Moving house mobility
-.042 -.030 -.025
(.046) (.047) (.047)
Gini coefficient
-.003 -.004
(.022) (.022)
% violent crimes
.051 .054
(.050) (.050)
% non-violent crimes
-.085 -.088
(.059) (.059)
Ethnicity neighbor(s) different
.034***
(.008)
cons
.687*** .675*** .489*** .490*** .507*** .510***
(.011) (.011) (.022) (.054) (.056) (.056)
Log-likelihood
-211.4 559.7 737.8 743.7 746.2 756.4
Source: SPVA 1998, Dutch Bureau of
Statistics
* p<0.05, ** p<0.01, ***
p<0.001
Page 25
24
Table 5. Multilevel logistic regression predicting neighbor with a different ethnicity or not,
Nindiv=5,757; Nzipcode= 291; Nmun= 13 (odds ratios, standard errors between brackets).
Model 1 Model 2 Model 3 Model 4 Model 5
Herfindahl Index of Ethnic Diversity 1.760* .104*** .116*** .457
.633
(.390) (.030)
(.034)
(.249)
(.355)
Dutch
ref.
ref.
ref.
ref.
Turkish
77.332*** 118.351*** 118.559*** 119.305***
(13.248) (25.624) (25.755) (25.909)
Moroccan
117.938*** 169.175*** 169.736*** 170.282***
(20.654) (36.419) (36.628) (36.721)
Surinamese
136.959*** 165.026*** 165.292*** 167.770***
(23.587) (29.891) (30.007) (30.459)
Antillean
523.427*** 652.465*** 663.091*** 678.834***
(109.169) (145.572) (148.232) (151.857)
Second generation
1.037
.962
.953
.959
(.135)
(.136)
(.135)
(.136)
Female
.743**
.735*** .740***
(.067)
(.067)
(.067)
Age
1.384
1.329
1.373
(.353)
(.339)
(.351)
Married
.787*
.788*
.789*
(.076)
(.077)
(.077)
Primary education
ref.
ref.
ref.
Lower secondary
.921
.923
.922
(.099)
(.100)
(.100)
Upper secondary
1.156
1.142
1.149
(.133)
(.131)
(.132)
College/University
1.444*
1.406*
1.408*
(.224)
(.219)
(.220)
No info on educational level
1.196
1.192
1.200
(.282)
(.280)
(.282)
Language proficiency
1.319
1.309
1.338
(.225)
(.224)
(.229)
Dutch citizen
1.076
1.074
1.084
(.110)
(.110)
(.111)
Family income
.840
.811
.789
(.399)
(.385)
(.375)
Dummy imputation family income
.974
.966
.955
(.104)
(.103)
(.102)
Home ownership
1.056
1.082
1.074
(.132)
(.138)
(.137)
Satisfaction with current finances
1.233
1.223
1.225
(.200)
(.199)
(.199)
% in higher education
3.255*
2.163
(1.672)
(1.136)
Mean income per person
5.071
5.856
(4.890)
(5.564)
Inequality ratio
1.352
1.373
(.552)
(.556)
Population density
.634
.803
(.226)
(.293)
Page 26
25
% over 65 years old
1.512
1.870
(.697)
(.874)
% renters
.670
.614
(.282)
(.261)
Moving house mobility
.222*
.289*
(.136)
(.178)
Gini coefficient
1.553
(.486)
% violent crimes
.304
(.207)
% non-violent crimes
2.262
(1.837)
Constant
2.097*** .018*** .011*** .015*** .013***
(.179) (.003)
(.004)
(.012)
(.010)
Log-likelihood
-3734.4 -2458.8
-2440.7
-2433.4
-2428.1
Source: SPVA 1998, Dutch Bureau of Statistics
* p<0.05, ** p<0.01, *** p<0.001
Page 27
26
Appendix
Construction of the dependent variables using cumulative scaling
For the measurement of trust we make use of Item Response Theory (IRT). IRT is used for
modeling latent traits and is based on the pattern in the items regarding the number of people that
gave a positive response. By taking into account the ordinal structure of the data, one constructs a
scale that is not purely based on the correlation between the items. Since social trust is often
understood in terms of ‘more’ and ‘less’ IRT, is especially suitable for measurement. It may
therefore be more appropriate for scale construction than for example factor analysis (Van Schuur &
Kiers, 2004).
Therefore a non-parametric IRT model for finding cumulative scales is used, the so-called
‘Mokken scaling method’. The software used is the Mokken Scale Analysis for Polytomous Items,
MSPWIN 5.0 (Molenaar & Sijtsma, 2000).This resulted in three scales (see tables A1, A3, A5).
First, a scale that measures ‘trust in neighbors’; containing items that deal with the intensity, quality
and frequency of contact with the direct neighbors of the respondent. Second a scale that measures
‘trust in the neighborhood’. Third, a scale that measures social distance, or ‘inter-ethnic trust’.
There are several criteria that a set of items has to meet to form an acceptable Mokken scale.
The relevant coefficients are presented in tables A2, A4 and A6, both for the total sample and split
out for the different ethnic groups. The most important measure is Loevinger’s Homogeneity
coefficient (H). The following cut-off values are conventional to judge a Mokken scale: >.30 being
a useful scale, >.40 a medium strong scale, and >.50 a strong scale (Mokken, 1996; Van Schuur,
2003). For each of the scales, H >.4. Furthermore, the test for monotone homogeneity (i.e. the
positive response to each item is a function of the positive response to easier items in the same
scale) and double monotonicity (to assess whether the degree of difficulty across items is the same
for all individuals) is positive. Last, the Cronbachs Alpha for the scales is satisfactory.
The actual scale consists of the sum of the items, and is than set to vary between 0 and 1.
Before this computation, missing values for the individual items were imputed using two-way
imputation (described in Sijtsma & Van der Ark, 2003). The imputation is done as follows (Van
Ginkel & Van der Ark, 2007, p. 2): ‘Let PMi be the average of all observed scores of respondent i,
let IMj be the average of all observed scores on item j, and let OM be the average of all observed
scores on all items and all persons. The missing value of respondent i on item j is then based on Xij
= PMi + IMj − OM’. Imputation was done for all cases with less than 60% of the scale items
missing. Those cases with more than 60% of the values missing were deleted.
Table A1. Items scale ‘trust in neighbors’.
Trust in neighbors
Mean
Range
Frequency of contact with one neighbor
2.18
1-4
Frequency of contact with other neighbor
2.10
1-4
Quality of contact with one neighbor
3.83
1-5
Quality of contact with other neighbor
3.79
1-5
Opinion on moving of one neighbor
3.43
1-5
Opinion on moving of other neighbor
3.38
1-5
Degree of nuisance with one neighbor (item
reversed)
2.82
1-3
Degree of nuisance with other neighbor (item
reversed)
2.82
1-3
Source: SPVA 1998.
Table A2. Scale values and coefficients ‘trust in neighbors’.
Page 28
27
Scale value (0-
1)
Homogeneity coefficient
(H)
Cronbachs
Alpha
Turks
.59
.46
.77
Moroccans
.59
.56
.78
Surinamese
.58
.42
.74
Antilleans
.57
.43
.77
Dutch
.61
.37
.70
Total
.59
.44
.75
Source: SPVA 1998.
Page 29
28
Table A3. Items scale ‘trust in neighborhood’.
Trust in neighborhood
Mean
Range
Frequency of contact among people in the
neighborhood
2.35
1-4
Feels comfortable in this neighborhood
3.35
1-4
Misses people in neighborhood when they move
1.44
1-3
Quality of contact among people in the
neighborhood
3.56
1-5
Source: SPVA 1998.
Table A4. Scale coefficients ‘trust in neighborhood’.
Scale value
Homogeneity
coefficient (H)
Cronbachs Alpha
Turks
.65
.40
.61
Moroccans
.61
.42
.70
Surinamese
.62
.46
.64
Antilleans
.60
.47
.62
Dutch
.65
.49
.65
Total
.63
.45
.64
Source:
SPVA 1998.
Table A5. Items scale ‘inter-ethnic trust’.
Mean
Range
Opinion on ethnic background friends of children
4.07
1-5
Opinion on ethnic background partner of children
3.68
1-5
Source: SPVA 1998.
Table A6. Scale coefficients ‘inter-ethnic trust’.
Scale value
Homogeneity
coefficient (H)
Cronbachs Alpha
Turks
.56
.60
.69
Moroccans
.57
.48
.61
Surinamese
.85
.67
.74
Antilleans
.88
.71
.88
Dutch
.69
.63
.79
Total
.72
.68
.76
Source:
SPVA 1998.
Page 30
29
Table A7. Correlation between the three indicators of individual trust (Significance level
between parentheses).
Trust in
direct
Trust in the Interethnic
trust
Trust in the
neighborhood
0.50
(0.00)
1
Interethnic trust
0.00
(0.94)
-0.00
(0.79)
1
Source: SPVA 1998.
Construction of the independent variable language proficiency
In table A8 and A9, the items and relevant coefficients for the language proficiency scale are
presented. The procedure followed is equal to that of the construction of the trust scales.
Table A8. Items scale language proficiency.
Mean
Range
Problems with reading Dutch (item inversed)
2.10
1-3
Frequency of using Dutch with partner
1.88
1-3
Frequency of using Dutch with children
2.17
1-3
Problems with speaking Dutch (item inversed)
2.39
1-3
Source: SPVA 1998.
Table A9. Scale coefficients language proficiency.
Scale value
Homogeneity
coefficient (H)
Cronbachs Alpha
Turks
.35
.60
.68
Moroccans
.42
.69
.70
Surinamese
.84
.54
.57
Antilleans
.77
.46
.56
Total
.63
.74
.74
Source:
SPVA 1998.