Emerging Trends in The Social and Behavioral Sciences

Social Network Analysis in the Study of Ethnic Inequalities

Media

Part of Social Network Analysis in the Study of Ethnic Inequalities

Title

Social Network Analysis in the Study of Ethnic Inequalities

extracted text

Social Network Analysis in the Study
of Ethnic Inequalities
FRANK KALTER

Abstract
Standard large-scale survey designs and methods enabled integration research to
progress far in recent decades, emphasizing especially the structural aspects of ethnic minorities’ integration. To further increase our understanding, the role social
aspects play in the complex process of integration merits more attention. Within this
endeavor, network analytical designs and techniques provide a particularly promising complement to the standard empirical research agenda. Network analysis provides adequate measures for diverse subaspects of social integration and allows to
tackle key open questions and issues, such as disentangling mechanisms of choice
from those of opportunity structure or of selection from influence. The use of network
analytical tools in integration research corresponds to the more general program of
analytical sociology calling for a stronger weight of contexts and social interactions
within the next generation of empirical research. While standard survey designs and
data sets study integration processes pretty much as if actors behaved in isolation,
integration is actually a heavily interactive and highly complex dynamic process.

THE ROLE OF SOCIAL INTEGRATION AND SOCIAL INTERACTIONS
IN THE STUDY OF ETHNIC INEQUALITIES
Research on the integration of immigrants and their descendants increased
tremendously in recent decades. This is not astonishing, given the rising
shares of the population with a migration background in almost all economically highly developed countries, and the societal, political, and scientific
challenges societies’ growing ethnic diversity entails. These challenges
pertain to a multitude of diverse aspects: language, education, economic
well-being, participation, identification, attitudes, values, and many more.
Accordingly, integration research has long stressed that “integration” is not
a monolithic concept, but one that needs to be broken down into different
dimensions (Alba & Nee, 1997). It is helpful, and meanwhile common, to
distinguish, at least on a very broad level, between structural integration,
Emerging Trends in the Social and Behavioral Sciences.
Robert Scott and Marlis Buchmann (General Editors) with Stephen Kosslyn (Consulting Editor).
© 2016 John Wiley & Sons, Inc. ISBN 978-1-118-90077-2.

1

2

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

social integration, and cultural integration—each dimension consisting of
many more fine-grained subaspects.
Notwithstanding this shared emphasis on the multifaceted nature of
integration, the bulk of mainstream quantitative research so far clearly falls
into the domain of structural integration, comprising countless larger-scale
studies on the integration of ethnic minorities into the labor markets and
the educational systems. The strong focus on these structural aspects is well
justified, as education and work are core areas, providing central resources
such as knowledge, skills, status, and income, which are important for many
further aspects of life; hence, the structural side is theoretically widely seen
as the key dimension of integration. Classical assimilation theory (Gordon,
1964) essentially assumes that once structural integration is achieved most
social and cultural aspects of integration will follow as a consequence. The
underlying theoretical argument basically being that structural integration
provides the opportunity structure necessary for the other integration
processes to succeed.
Large-scale empirical research has advanced far in detecting important patterns of structural ethnic inequalities and in understanding major mechanisms behind them. One of the main general insights certainly is that ethnic
inequalities are to a large extent actually class inequalities in disguise. In
particular, comparative quantitative research reveals that for most groups in
most countries, ethnic differences in educational attainment are a matter of
socioeconomic status (SES) (Heath & Brinbaum, 2014), and ethnic differences
in occupational attainment are to a great degree a matter of education (Heath
& Cheun, 2007)—thus confirming basic human capital theory arguments.
However, while SES can usually explain the lion’s share of group and country differences, it cannot tell the whole story. Even controlling for SES one
finds notable effects of ethnic group membership on educational success, and
even controlling for education one finds these same effects on labor market
outcomes. These residual direct effects of ethnic group membership are the
center of interest in current research on ethnic structural inequalities.
To explain these ethnic residuals very different types of theoretical reasoning have evolved that await stricter empirical testing (Heath, Rothon, &
Kilpi, 2008; Kalter, 2011). Here, at the latest, the importance of social integration reenters the scene. Next to arguments referring to aspects of cultural integration, many of the potential mechanisms suggested can roughly
be subsumed under the general idea that how minorities are embedded in
social relations and interactions might be responsible, and that these social
aspects are not merely a consequence, but rather, at least partly, also a cause
of structural integration. Behind the major lines of reasoning, two more general approaches have become especially prevalent. The first is social capital

Social Network Analysis in the Study of Ethnic Inequalities

3

theory, according to which ethnic residuals might have to do with the equipment with resources that are available via social ties and networks (Portes,
1995). The basic idea builds upon the seminal work of Granovetter (1973) on
the role of social contacts for labor market success, expanding it also to other
structural areas such as educational attainment (Coleman, 1988) and focusing specifically on the relative roles of coethnic versus interethnic ties, often
also called bonding versus bridging social capital (Putnam, 2007). A second set
of frequent arguments refers to several variants of theories of discrimination
(Blank, Dabady, & Citro, 2004, pp. 55–70).
These frameworks suggest that greater and more systematic attention must
be paid to the “social” dimension of integration. However, the means to do
so are limited given the overall data infrastructure for quantitative empirical research. Integration research has largely benefited from an increased
availability of microdata from the official statistics and from a number of
large-scale social surveys that provide a sufficiently large overall sample size
to study minorities. As a rule, these bigger data sets contain ample information on respondents’ educational, occupational, and general economic situation. Thus, they cover especially the structural dimension of integration very
well, which—next to the theoretical reasons—might further explain their so
far dominant role in research. Unfortunately, however, the large-scale surveys commonly used for the empirical analysis of structural ethnic inequalities often contain no information on aspects of social integration.
Those that included respective measures clearly seem able to confirm
that social integration might indeed be an important piece of the puzzle.
An example is the German Socioeconomic Panel (GSOEP), which has
intensively been employed to study the impact of contacts and friendship
ties (to coethnics and natives) on the labor market integration of the classical
immigrant groups in Germany (Kalter, 2011; Kanas, Chiswick, van der
Lippe, & van Tubergen, 2012; Lancee, 2012). Other prominent examples
are the Mexican Migration Project (MMP, Aguilera & Massey, 2003); the
Children of Immigrants Longitudinal Study (CILS, Portes & MacLeod, 1999);
or the Dutch Social Position and Use of Welfare Facilities by Immigrants
survey (SPVA, Lancee, 2010).
However, even if measures are available, the means of standard survey
research to address key open questions of social integration by design remain
severely restricted. Against this background, network analytical studies are a
necessary and especially promising complement to standard survey designs.
Network analytical tools have a long history and well-proven utility in the
social sciences; they immediately correspond to the central ideas around the
concepts of social integration and social interactions. Moreover, the development of new techniques and models that provide new opportunities to

4

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

contribute to the key open questions carved out to date by empirical integration research has progressed tremendously.
In the following, we seek to illustrate with selective examples rather than
in a systematic manner how the network analytical approach might contribute. In the first section, we show how it helps in deriving theoretically
more meaningful measures of many important subaspects of the concept
of social integration. In the second section, we outline how it helps in better detecting the causes behind social integration, not least because of its
unique allowance to adequately control the opportunity structure and “natural” dynamic interaction processes. In the third section, we show how new
techniques can perhaps help to tackle the fundamental problem of disentangling the exact causal interplay between social and other important aspects
of integration. We conclude by putting the ideas and potentials into a more
general setting, arguing that the use of new network analytical tools in integration research can be subsumed under the more general program of analytical sociology that emphasizes the role social relations and social interactions
in the explanation of social phenomena.
TOWARD BETTER MEASURES OF WHAT SOCIAL INTEGRATION
REALLY MEANS
As mentioned above, measures of ethnic minorities’ social integration have
been applied quite successfully in standard survey designs. Among the
particular instruments used, ego-centered network questions are the most
sophisticated. Respondents are asked to name a number of specified ties,
for example, the three best friends, and to report certain characteristics of
these ties. This information is then used to construct variables of interest
for the analyses; here, the relative share of majority members or coethnics
among the ties is the most frequently used indicator. Though relatively
time-expensive, the approach is in principle quite flexible, and can, for
example, target many different resources of the named persons so as to very
narrowly capture the concept of relevant social capital.
Within an explicit network design, however, sociometric questions ask
a respondent whether there are certain specified relations to each of the
other respondents. This then leads to a complete network specific for the
respective relation. Figure 1, for example, shows a friendship network for a
German classroom within the Children of Immigrants Longitudinal Survey
in Four European countries (CILS4EU, Kalter et al., 2015). It results from
the sociometric question:” Who are your best friends in class? “The colors
denote different ethnic backgrounds (yellow, majority without migration
background; green, Turkish background; and red, any other background).
The size of a dot represents the “strength” of a migration background (large,

Social Network Analysis in the Study of Ethnic Inequalities

5

8
14
9

2

7
4

18

11

15

6

3

10

13

16

17

Figure 1

1

5

12

Example of a classroom friendship network from the CILS4EU data.

born abroad; medium, at least one parent born abroad; and small, both
parents native-born). Arrows represent friendship nominations, pointing
from a nominator to the nominee.
Figure 1 shows, already intuitively, that the sociometric approach contains
very rich information about the social integration in the classroom. In general, it bears many well-known comparative advantages. In the following,
we stress four points that seem particularly important in the context of ethnic
minority research.
The first advantage is obviously that the information on the ties stems from
interviews with these persons themselves and is thus much more comprehensive and reliable. In integration research, the benefits of this fact already
start with the very concept of ethnicity itself. The evidence of a person’s
social integration as indicated by the share of minority members among the
nominated friends might differ according to the definition of ethnic minority
groups. In Figure 1, for example, students 6 and 14 would have equal value
(100%) when applying a wide concept. However, student 6 would have a
value lower than 14, if student 1, who belongs to the third generation, would
count as “majority” in a more narrow definition of ethnic minorities. In an

6

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

ego-centered approach, it seems almost impossible for a respondent to know
all the relevant details about his social ties needed to classify their ethnicity
correctly, for example, whether this person’s grandparents were born abroad,
or—as is common in the United Kingdom—someone subjectively identifies
with a certain group or not. Moving from the pure ethnicity of the social ties
to the social capital they include, the issues continue. This holds especially
when “softer” characteristics become theoretically relevant aspects, which is
very much the case within integration-related research questions—language
proficiency, cultural knowledge, or national identification of the social ties
being obvious examples.
The second, closely related point is that integration in general and social
integration in specific are not one-sided processes (Alba & Nee, 1997), but
heavily interactive. Information on how others, for example, majority classmates, treat a minority student at interest is obviously an important part of
the story. Looking back at Figure 1, the fact that student 6 is nominated as a
friend by a majority student (10) makes him, in a specific sense, somewhat
better integrated than student 14. Note that having information on the behavior and preferences of others also opens a door to measuring and studying
processes of discrimination in fresh new ways. This holds true all the more,
as the network approach allows us to examine very different kinds of relations between any two persons, among them so-called negative ties such as
disliking or bullying (Tolsma, van Deurzen, Stark, & Veenstra, 2013). Many
surveys still try to grasp discrimination by including questions on perceived
discrimination, despite long-standing doubts about the validity of these measures and thus the potential of drawing any causal conclusions (Blank et al.,
2004, p. 16ff).
The third advantage of the sociometric approach is that, casually speaking, it allows to study social integration at “higher orders”. Looking again at
our two exemplary students in Figure 1, student 6 is better socially integrated
than student 14 in the sense that he is befriended by students who themselves
have relatively many friends, among them majority members. Furthermore,
metaphorically speaking, overall he seems to be better “placed” within the
classroom than student 14. Social network analysis provides us now with a
rich repertoire of different measures that capture more precisely the general
position of an individual within a given network, for example, reachability,
actor centrality, or actor prestige (Wassermann & Faust, 1994). Such measures exist in many subvariants and can potentially be fine-tuned to more
adequately capture theoretically relevant aspects of social integration.
Fourth, and very importantly, the measures of social integration are not
restricted to being individual characteristics only. On the contrary, a major
attraction of network analysis is that it has developed many helpful measures describing characteristics of the network as a whole, such as density,

Social Network Analysis in the Study of Ethnic Inequalities

7

cliquishness, cohesiveness, and amount of clustering (Jackson, 2008, p. 20ff).
Many questions and discussions in integration research are related to concepts of social integration that are basically better understood as properties
of larger social units rather than of individual actors. A good example is the
general debate about the potential effects of ethnic diversity on social cohesion, stimulated by a meanwhile famous article by Robert Putnam (2007). The
empirical research on this topic typically employs individual-level survey
data using generalized trust, reported civic engagement, attitudes toward
the welfare states, or similar variables as indicators of social cohesion. Social
cohesion, however, is not only a psychological and attitudinal concept, but
importantly and foremost consists of relational aspects.
How ethnic diversity and social cohesion are empirically related thus
can and should also be addressed taking a bird’s eye view. For example,
in the network in Figure 1, the students in the classroom seem fairly well
connected. This can be specified more precisely in network analytical terms:
a very simple measure, for example, would be network density, which is
defined as the number of actual ties in the network divided by the number
of all ties possible. The value for the network in figure would be 0.18. One
can also attach numerical values to the ethnic diversity of the classroom.
Applying the meanwhile common Herfindahl-based index, the value for
the classroom in Figure 1 is 0.84. On the basis of this logic, Kalter and Kruse
(2015) show that in the representative sample of 184 classroom networks
in Germany within the CILS4EU study, there is no correlation between
ethnic diversity and network density. They also find similar results for the
representative classroom samples in England, the Netherlands, and Sweden.
Findings change only slightly upon application of more sophisticated indices
of social cohesion. The example thus shows that the network analytical
approach delivers a fresh, and in some respects more adequate empirical
view of important long-standing questions in integration research.
TOWARD THE MECHANISMS BEHIND SOCIAL INTEGRATION
If social integration is an important piece of the ethnic inequality puzzle, then
the factors accounting for it move into the focus of integration research. In
particular, thinking back to the above-mentioned idea of classical assimilation theory—it is important to analyze whether social integration is more
than just a mere consequence of the opportunity structure deriving from
structural integration, and, if so, why. Here, recently developed multivariate
models in network analysis meanwhile provide a rich toolbox of techniques.
One of the techniques that has already received considerable attention
for ethnic minority-related topics is Exponential Random Graph Models
(ERGM) or, synonymously, p* models (Robins, Pattison, Kalish, & Lusher,

8

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

2007). Basically, these apply a logistic regression approach to dyads of
network, thus allowing to assess which factors influence whether any two
persons within such a network have a certain relation or not. The explaining
factors can include characteristics of the individuals on either ends of a tie as
well as characteristics of the dyad itself, for example, homogeneity of both
individuals with respect to certain characteristics, whether they share other
ties, and more complex structural features.
This has accounted for much recent progress, especially in the study of
ethnic or racial homophily (Kruse, Smith, van Tubergen, & Maas, 2016;
Moody, 2001; Mouw & Entwisle, 2006; Stark, 2011; Wimmer & Lewis, 2010).
The finding is almost universal that friendships tend to be strongly ethnically
homogenous, which is often interpreted as an explicit preference of ethnic
group members: the unwillingness to integrate socially, on the part of the
minorities, or social discrimination, on the part of the majority. However,
the reasons for this can be very different and rather diverse and are hard
to disentangle using standard survey methods. First of all, it could simply
be a matter of opportunity structure. In this respect, the network analytical
approach in general and the ERGMs in particular allow a uniquely adequate
control, as the pool from which friends can be chosen at all is specified by the
complete network and, casually spoken, enters the denominator of the effect
estimation. Second, ethnic homogeneity can also result from preferences for
or the impact of characteristics that are only related to ethnicity, but should
not be confused with an explicit preference for ethnicity itself. For example,
students might prefer to befriend students from a similar social class background, as this means a similar economic situation of their families, allowing
them to engage in similar leisure activities. Similarly, additional and less
obvious structural arrangements might foster the ethnic homogeneity of
ties; students from the same ethnicity might for several reasons be more
likely to share different tracks within schools and extracurricular activities
or come from the same neighborhood. ERGM allows us to control for all this.
Finally, important built-in network mechanisms, sometimes called balancing mechanisms, can amplify the ethnic homogeneity of social ties, thus overestimating potentially underlying preference effects. Most basically, friendship choices and other sorts of social relations tend to be reciprocated. So
even if A chooses B because of an ethnic preference, the fact that B chooses
A might be due to simple reciprocity and itself not reveal any ethnic preference. Similarly, triads tend to closure, so if A and B have a relation and B and
C have a relation, it is very likely that A and C will also have a relation. It is
a strategic feature of ERGMs that they very conveniently and flexibly allow
to control for these kinds of “natural” network mechanisms when assessing
tendencies for ethnic friendship choices.

Social Network Analysis in the Study of Ethnic Inequalities

9

TOWARD DISENTANGLING THE COMPLEX RELATION BETWEEN
DIFFERENT ASPECTS OF INTEGRATION—SELECTION VERSUS
INFLUENCE
The network analytical approach also gives us a better methodological grip
on a key issue in integration research: the direction of causality between different aspects of integration. As sketched above, the literature suggests that
many mechanisms that consider certain aspects of social integration lead to
structural ethnic inequalities, such as the Granovetter type of social capital
effects on labor market outcomes. However, as also noted with reference to
classical assimilation theory, arguments in the other direction are at least
likewise plausible. Obviously, the answer to what is the right direction is
of major importance, not least for any integration-related policies and measures. Here, studying networks in longitudinal perspective offers some new
strategic tools of analysis that promise much better insights. They do so by
enabling us to empirically disentangle the two important general mechanisms of selection and influence.
Let us assume, for example, that we are dealing with two ethnic
groups—blues and yellows—and that blues on average do worse in school
than yellows. Let us also assume that, at a certain point in time (t), those
blues who perform better have relatively more yellows as friends. (If we
think of blues as representing an ethnic minority and yellows as representing
the majority this is empirically a quite frequent scenario.) This situation is
illustrated by some small networks in the right column (time t) in Figure 2,
where the size of a dot expresses school performance and lines represent
friendships. Basically, this relation could result from the fact that for the
yellows those blues who do better are more attractive choices as friends
and/or that the more successful blues are more likely to select yellows
as friends (because these are more likely good performers). This would
represent mechanisms of selection as illustrated by the transition from time
t − 1 to time t in Figure 2, where the focus is always on the blue in the middle.
It could also be, however, that those blues who are more strongly befriended
by yellows (for whatever reason) at time t − 1 profit from these ties and
become better performers at time t. This would correspond to the selection
mechanisms sketched in Figure 2.
Figure 2 thus shows that similar constellations at time t can arise from different starting points at time t − 1 via the different mechanisms of selection
versus influence, and that similar starting points at t − 1 can lead to different constellations at time t. Comparing networks over time thus allows to
distinguish selection effects from influence effects.
Meanwhile, sophisticated statistical techniques allow these ideas to be
applied to empirical network panel data. Tom Snijders and his workgroup

10

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Time t−1

Time t

on

ecti

Sel

Influ

enc

e

ion

ect

Sel

Infl

uen

ce

Figure 2 Illustration of mechanisms of selection and influence in small networks;
focus is on the blue dot in the middle.

developed stochastic actor-oriented models (SAOM) for network dynamics
(Snijders, van de Bunt, & Steglich, 2010) that permit uniquely strict tests.
Basically, these are agent-based simulation models assuming that actors
make choices about changing their social ties and their behavior. The decision
to form or cut specific ties and the decision for or against a specific change
in behavior are assumed to be dependent on a set of variables—pretty much
like in a regression model. Without going further into technical details of the
complex and assumption-rich estimation procedure, the important thing is
that this yields coefficient estimates and estimates for their standard errors
both for variables related to the network dynamics and, simultaneously, for
variables related to the behavior dynamics—among them coefficients that
directly capture the selection and influence mechanisms. Estimation can be
conducted with a software program called SIENA (Simulation Investigation
for Empirical Network Analysis), which is meanwhile integrated in the general
software package R and in this form also known under the name RSiena
(Ripley, Snijders, Boda, Vorce, & Preciado, 2015).
That SAOM techniques are meanwhile well established in many fields
is proved by the long list of published applications on the SIENA website

Social Network Analysis in the Study of Ethnic Inequalities

11

(www.stats.ox.ac.uk/∼snijders/siena/). A frequently researched topic,
for example, is the coevolution of adolescents’ friendship networks and
different forms of deviant behavior or delinquency, such as nicotine, drug,
and alcohol consumption, or carrying a weapon. With the rising shares of
ethnic minority students among the youth in almost all countries, variables
of ethnicity have occasionally also attracted interest, not only among the
controls. In the course of the vivid activities and progress, researchers are
increasingly aware of the potential of SAOM tools for addressing classical
questions of integration research, and some have already targeted these
more explicitly.
A recent example is the paper by Leszczensky, Stark, Flache, and
Munniksma (2015), which uses the data of the Arnhem School Study (TASS,
see Stark, 2011) to study the coevolution of friendship networks and host
country identification, thus disentangling the causal relationship between
aspects of social integration and cultural integration. Although there is an
empirical correlation between the strength of identification with the host
country and the number of native friends among adolescent immigrants,
the study does not support the hypothesis that influence mechanisms are
at work. On the selection side, there is also no evidence for the assumption
that immigrant youth who identify more strongly with the Netherlands are
more likely to prefer native-born Dutch friends. Interestingly, the analyses
show that, in turn, native-born Dutch significantly prefer immigrants who
identify more strongly with the Netherlands. The study is thus an especially
telling proof that in order to correctly understand the mechanisms behind
integration processes, it is important to consider relational aspects. While
the example focuses on the interplay between social and cultural aspects of
integration, it is easy to imagine similar interesting applications targeting
the relation between ethnic minorities’ social integration and structural
aspects, such as school performance and truancy.
NETWORK ANALYSIS, INTEGRATION RESEARCH, AND THE
PROGRAM OF ANALYTICAL SOCIOLOGY
In this essay, we have outlined the potential of social network analysis and
some new developments in that field for the study of ethnic inequalities. In
particular, we have exemplified how they can increase our understanding
of the precise role of social aspects in the complex process of integration.
Network analytical designs and techniques represent a promising complement to the survey data-based standard empirical research agenda; they provide particular adequate measures for diverse subaspects of the concept of
social integration and the recent development of sophisticated statistical network methods allows to tackle key open questions and issues of integration

12

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

research, such as disentangling the mechanisms of choice versus those of
opportunity structure or of selection versus influence.
The growing awareness and use of new network analytical tools in integration research is well in line with a more general theoretical movement and
orientation that has recently become known and more visible under the label
“analytical sociology” (Hedström & Bearman, 2009). Next to more epistemological messages emphasizing the general need for truly mechanism-based
explanations, a major impetus of this approach lies in the fact that—while
doubtless standing on the firm ground of methodological individualism—it
puts more attention and relative weight on the impact of social contexts and
social interactions rather than on all too deep details of action theories (Kalter
& Kroneberg, 2014).
Large-scale research survey data analysis has been a huge success story in
many fields, due not least to the fact that it can be guided by explicit and
elaborated theories of action (Kroneberg & Kalter, 2012). Human capital theory has always played a dominant role in migration and integration research
and, as briefly sketched at the beginning, is as a rule able to account for
much of what is going on in fields such as the labor market or education.
Despite all the progress, however, a major limitation of quantitative empirical ethnic inequality research has been that standard survey designs and
data sets force us to assume that actors behave independently in isolation
(Snijders & Steglich, 2015). Integration, however, is—like all interesting social
phenomena—a heavily interactive and highly complex dynamic process, in
which the actions of one actor immediately constitute an important context
and condition for the actions of the other, and so forth. The next generation of
empirical research must complement the traditional individual survey data
approach by designs and strategies of analysis that take these contextual and
interactional processes into explicit account. For many key open questions,
longitudinal network analytical designs seem to be among the most promising of these tools.
REFERENCES
Aguilera, M. B., & Massey, D. S. (2003). Social capital and the wages of Mexican
migrants: New hypotheses and tests. Social Forces, 82(2), 671–701.
Alba, R., & Nee, V. (1997). Rethinking assimilation theory for a new era of immigration. International Migration Review, 31(4), 826–874.
Blank, R. M., Dabady, M., & Citro, C. F. (Eds.) (2004). Measuring racial discrimination.
Washington, D.C.: National Academies Press.
Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal
of Sociology, 94(Supplement 95), 95–120.
Gordon, M. M. (1964). Assimilation in American life: The role of race, religion and national
origins. New York, NY: Oxford University Press.

Social Network Analysis in the Study of Ethnic Inequalities

13

Granovetter, M. S. (1973). The strength of weak ties. American Journal of Sociology,
8(6), 1360–1380.
Heath, A., & Brinbaum, Y. (Eds.) (2014). Unequal attainments. Ethnic educational
inequalities in ten western countries. Oxford, England: British Academy; Oxford University Press.
Heath, A., & Cheun, S. Y. (Eds.) (2007). Unequal chances. Ethnic minorities in western
labour market. Oxford, England: British Academy; Oxford University Press.
Heath, A. F., Rothon, C., & Kilpi, E. (2008). The second generation in Western Europe:
Education, unemployment, and occupational attainment. Annual Review of Sociology, 34, 211–235.
Hedström, P., & Bearman, P. (Eds.) (2009). The Oxford handbook of analytical sociology.
New York, NY: Oxford University Press.
Jackson, M. O. (2008). Social and economic networks. Princeton, NJ: Princeton University Press.
Kalter, F. (2011). The second generation in the German labor market. In R. Alba & M.
C. Waters (Eds.), The next generation: Immigrant youth in a comparative perspective.
New York: NYU Press.
Kalter, F., Heath, A., Hewstone, M., Jonsson, J. O., Kalmijn, M., Kogan, I., & van
Tubergen, F. (2015). Children of immigrants longitudinal survey in four European
countries (CILS4EU). GESIS Data Archive, Cologne. ZA5353 Data file Version
2.1.0, doi: 10.4232/cils4eu.5353.2.1.0
Kalter, F., & Kroneberg, C. (2014). Between mechanism talk and mechanism cult:
New emphases in explanatory sociology and empirical research. In J. Friedrichs
& A. Nonnenmacher (Eds.), Kölner Zeitschrift für Soziologie und Sozialpsychologie,
Special Issue 54: Social Contexts and Social Mechanisms (pp. 91–115). Wiesbaden, Germany: Springer VS.
Kalter, F., & Kruse, H. (2015). Ethnic diversity, homophily, and network cohesion
in European classrooms. In R. Koopmans, B. Lancee & M. Schaeffer (Eds.), Social
cohesion and immigration in Europe and North America: mechanisms, conditions, and
causality. London, England: Routledge.
Kanas, A., Chiswick, B. R., van der Lippe, T., & van Tubergen, F. (2012). Social contacts and the economic performance of immigrants: A panel study of immigrants
in Germany. International Migration Review, 46(3), 680–709.
Kroneberg, C., & Kalter, F. (2012). Rational choice theory and empirical research.
Methodological and theoretical contributions in Europe. Annual Review of Sociology, 38, 73–92.
Kruse, H., Smith, S., van Tubergen, F., & Maas, I. (2016). From neighbours to school
friends? How adolescents’ place of residence relates to same-ethnic school friendships. Social Networks, 44(1), 130–142.
Lancee, B. (2010). The economic returns of immigrants’ bonding and bridging social
capital: The case of the Netherlands. International Migration Review, 44(1), 202–226.
Lancee, B. (2012). The economic returns of bonding and bridging social capital for
immigrant men in Germany. Ethnic and Racial Studies, 35(4), 664–683.

14

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Leszczensky, L., Stark, T. H., Flache, A., & Munniksma, A. (2015). Disentangling
the relationship between young immigrants’ host country identification and their
friendships with natives. Social Networks, 44, 179–189.
Moody, J. (2001). Race, school integration, and friendship segregation in America.
American Journal of Sociology, 107(3), 679–716.
Mouw, T., & Entwisle, B. (2006). Residential segregation and interracial friendship
in schools. American Journal of Sociology, 112(2), 394–441.
Portes, A. (1995). Economic sociology and the sociology of immigration: A conceptual overview. In A. Portes (Ed.), The economic sociology of immigration (Vol. 29, pp.
1–41). New York, NY: Russell Sage Foundation.
Portes, A., & MacLeod, D. (1999). Educating the second generation: Determinants of
academic achievement among children of immigrants in the United States. Journal
of Ethnic and Migration Studies, 25(3), 373–396.
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.
Ripley, R.M., Snijders, T.A.B., Boda, Z., Vorce, A., & Preciado, P. (2015). Manual for RSiena. Retrieved from http://www.stats.ox.ac.uk/∼snijders/siena/
RSiena_Manual.pdf
Robins, G., Pattison, P., Kalish, Y., & Lusher, D. (2007). An introduction to exponential
random graph (p*) models for social networks. Social Networks, 29, 173–191.
Snijders, T. A. B., & Steglich, C. E. G. (2015). Representing micro–macro linkages by
sector-based dynamic network models. Sociological Methods & Research, 44, 22–271.
Snijders, T. A. B., van de Bunt, G. G., & Steglich, C. E. G. (2010). Introduction to
stochastic actor-based models for network dynamics. Social Networks, 32, 44–60.
Stark, T. H. (2011). Integration in schools. A process perspective on students’ interethnic
attitudes and interpersonal relationships (ICS dissertation). Groningen.
Tolsma, J., van Deurzen, J., Stark, T. H., & Veenstra, R. (2013). Who is bullying whom
in ethnically diverse primary schools? Exploring links between bullying, ethnicity,
and ethnic diversity in Dutch primary schools. Social Networks, 36, 51–61.
Wassermann, S., & Faust, K. (1994). Social network analysis: Methods and applications
(structural analysis in the social sciences (Vol. 8). Cambridge, England: Cambridge
University Press.
Wimmer, A., & Lewis, K. (2010). Beyond and below racial homophily: ERG models
of a friendship network documented on Facebook. American Journal of Sociology,
116, 583–642.

FRANK KALTER SHORT BIOGRAPHY
Frank Kalter is a professor of sociology at the School of Social Sciences at the
University of Mannheim, Germany. Currently, he also serves as the director
of the Mannheim Centre for European Social Research (MZES). He is a fellow
of the European Academy of Sociology (EAS) and has been its president from
2011 to 2015. His major research interests include migration, the integration of
immigrants and their children, and the formal modeling of social processes.

Social Network Analysis in the Study of Ethnic Inequalities

15

RELATED ESSAYS
Misinformation and How to Correct It (Psychology), John Cook et al.
Enduring Effects of Education (Sociology), Matthew Curry and Jennie E.
Brand
Cognitive Processes Involved in Stereotyping (Psychology), Susan T. Fiske
and Cydney H. Dupree
Controlling the Influence of Stereotypes on One’s Thoughts (Psychology),
Patrick S. Forscher and Patricia G. Devine
Culture and Social Networks (Sociology), Jan A. Fuhse
Migrant Networks (Sociology), Filiz Garip and Asad L. Asad
Ethnic Enclaves (Sociology), Steven J. Gold
Immigrant Children and the Transition to Adulthood (Sociology), Roberto G.
Gonzales and Benjamin J. Roth
Racial Disenfranchisement (Political Science), Vincent L. Hutchings and
Davin L. Phoenix
Emerging Trends in Social Network Analysis of Terrorism and Counterterrorism (Sociology), David Knoke
Immigration and the Changing Status of Asian Americans (Sociology),
Jennifer Lee
Immigrant Sociocultural Adaptation, Identification, and Belonging (Sociology), Sarah J. Mahler
Political Inequality (Sociology), Jeff Manza
Immigrant Health Paradox (Sociology), Kyriakos S. Markides and Sunshine
Rote
Politics of Immigration Policy (Political Science), Jeannette Money
The Role of School-Related Peers and Social Networks in Human Development (Political Science), Chandra Muller
Health and Social Inequality (Sociology), Bernice A. Pescosolido
Latinos and the Color Line (Sociology), Clara E. Rodríguez et al.
Digital Methods for Web Research (Methods), Richard Rogers
Stereotype Threat (Psychology), Toni Schmader and William M. Hall
Assimilation and Its Discontents (Sociology), Min Zhou



Social Network Analysis in the Study
of Ethnic Inequalities
FRANK KALTER

Abstract
Standard large-scale survey designs and methods enabled integration research to
progress far in recent decades, emphasizing especially the structural aspects of ethnic minorities’ integration. To further increase our understanding, the role social
aspects play in the complex process of integration merits more attention. Within this
endeavor, network analytical designs and techniques provide a particularly promising complement to the standard empirical research agenda. Network analysis provides adequate measures for diverse subaspects of social integration and allows to
tackle key open questions and issues, such as disentangling mechanisms of choice
from those of opportunity structure or of selection from influence. The use of network
analytical tools in integration research corresponds to the more general program of
analytical sociology calling for a stronger weight of contexts and social interactions
within the next generation of empirical research. While standard survey designs and
data sets study integration processes pretty much as if actors behaved in isolation,
integration is actually a heavily interactive and highly complex dynamic process.

THE ROLE OF SOCIAL INTEGRATION AND SOCIAL INTERACTIONS
IN THE STUDY OF ETHNIC INEQUALITIES
Research on the integration of immigrants and their descendants increased
tremendously in recent decades. This is not astonishing, given the rising
shares of the population with a migration background in almost all economically highly developed countries, and the societal, political, and scientific
challenges societies’ growing ethnic diversity entails. These challenges
pertain to a multitude of diverse aspects: language, education, economic
well-being, participation, identification, attitudes, values, and many more.
Accordingly, integration research has long stressed that “integration” is not
a monolithic concept, but one that needs to be broken down into different
dimensions (Alba & Nee, 1997). It is helpful, and meanwhile common, to
distinguish, at least on a very broad level, between structural integration,
Emerging Trends in the Social and Behavioral Sciences.
Robert Scott and Marlis Buchmann (General Editors) with Stephen Kosslyn (Consulting Editor).
© 2016 John Wiley & Sons, Inc. ISBN 978-1-118-90077-2.

1

2

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

social integration, and cultural integration—each dimension consisting of
many more fine-grained subaspects.
Notwithstanding this shared emphasis on the multifaceted nature of
integration, the bulk of mainstream quantitative research so far clearly falls
into the domain of structural integration, comprising countless larger-scale
studies on the integration of ethnic minorities into the labor markets and
the educational systems. The strong focus on these structural aspects is well
justified, as education and work are core areas, providing central resources
such as knowledge, skills, status, and income, which are important for many
further aspects of life; hence, the structural side is theoretically widely seen
as the key dimension of integration. Classical assimilation theory (Gordon,
1964) essentially assumes that once structural integration is achieved most
social and cultural aspects of integration will follow as a consequence. The
underlying theoretical argument basically being that structural integration
provides the opportunity structure necessary for the other integration
processes to succeed.
Large-scale empirical research has advanced far in detecting important patterns of structural ethnic inequalities and in understanding major mechanisms behind them. One of the main general insights certainly is that ethnic
inequalities are to a large extent actually class inequalities in disguise. In
particular, comparative quantitative research reveals that for most groups in
most countries, ethnic differences in educational attainment are a matter of
socioeconomic status (SES) (Heath & Brinbaum, 2014), and ethnic differences
in occupational attainment are to a great degree a matter of education (Heath
& Cheun, 2007)—thus confirming basic human capital theory arguments.
However, while SES can usually explain the lion’s share of group and country differences, it cannot tell the whole story. Even controlling for SES one
finds notable effects of ethnic group membership on educational success, and
even controlling for education one finds these same effects on labor market
outcomes. These residual direct effects of ethnic group membership are the
center of interest in current research on ethnic structural inequalities.
To explain these ethnic residuals very different types of theoretical reasoning have evolved that await stricter empirical testing (Heath, Rothon, &
Kilpi, 2008; Kalter, 2011). Here, at the latest, the importance of social integration reenters the scene. Next to arguments referring to aspects of cultural integration, many of the potential mechanisms suggested can roughly
be subsumed under the general idea that how minorities are embedded in
social relations and interactions might be responsible, and that these social
aspects are not merely a consequence, but rather, at least partly, also a cause
of structural integration. Behind the major lines of reasoning, two more general approaches have become especially prevalent. The first is social capital

Social Network Analysis in the Study of Ethnic Inequalities

3

theory, according to which ethnic residuals might have to do with the equipment with resources that are available via social ties and networks (Portes,
1995). The basic idea builds upon the seminal work of Granovetter (1973) on
the role of social contacts for labor market success, expanding it also to other
structural areas such as educational attainment (Coleman, 1988) and focusing specifically on the relative roles of coethnic versus interethnic ties, often
also called bonding versus bridging social capital (Putnam, 2007). A second set
of frequent arguments refers to several variants of theories of discrimination
(Blank, Dabady, & Citro, 2004, pp. 55–70).
These frameworks suggest that greater and more systematic attention must
be paid to the “social” dimension of integration. However, the means to do
so are limited given the overall data infrastructure for quantitative empirical research. Integration research has largely benefited from an increased
availability of microdata from the official statistics and from a number of
large-scale social surveys that provide a sufficiently large overall sample size
to study minorities. As a rule, these bigger data sets contain ample information on respondents’ educational, occupational, and general economic situation. Thus, they cover especially the structural dimension of integration very
well, which—next to the theoretical reasons—might further explain their so
far dominant role in research. Unfortunately, however, the large-scale surveys commonly used for the empirical analysis of structural ethnic inequalities often contain no information on aspects of social integration.
Those that included respective measures clearly seem able to confirm
that social integration might indeed be an important piece of the puzzle.
An example is the German Socioeconomic Panel (GSOEP), which has
intensively been employed to study the impact of contacts and friendship
ties (to coethnics and natives) on the labor market integration of the classical
immigrant groups in Germany (Kalter, 2011; Kanas, Chiswick, van der
Lippe, & van Tubergen, 2012; Lancee, 2012). Other prominent examples
are the Mexican Migration Project (MMP, Aguilera & Massey, 2003); the
Children of Immigrants Longitudinal Study (CILS, Portes & MacLeod, 1999);
or the Dutch Social Position and Use of Welfare Facilities by Immigrants
survey (SPVA, Lancee, 2010).
However, even if measures are available, the means of standard survey
research to address key open questions of social integration by design remain
severely restricted. Against this background, network analytical studies are a
necessary and especially promising complement to standard survey designs.
Network analytical tools have a long history and well-proven utility in the
social sciences; they immediately correspond to the central ideas around the
concepts of social integration and social interactions. Moreover, the development of new techniques and models that provide new opportunities to

4

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

contribute to the key open questions carved out to date by empirical integration research has progressed tremendously.
In the following, we seek to illustrate with selective examples rather than
in a systematic manner how the network analytical approach might contribute. In the first section, we show how it helps in deriving theoretically
more meaningful measures of many important subaspects of the concept
of social integration. In the second section, we outline how it helps in better detecting the causes behind social integration, not least because of its
unique allowance to adequately control the opportunity structure and “natural” dynamic interaction processes. In the third section, we show how new
techniques can perhaps help to tackle the fundamental problem of disentangling the exact causal interplay between social and other important aspects
of integration. We conclude by putting the ideas and potentials into a more
general setting, arguing that the use of new network analytical tools in integration research can be subsumed under the more general program of analytical sociology that emphasizes the role social relations and social interactions
in the explanation of social phenomena.
TOWARD BETTER MEASURES OF WHAT SOCIAL INTEGRATION
REALLY MEANS
As mentioned above, measures of ethnic minorities’ social integration have
been applied quite successfully in standard survey designs. Among the
particular instruments used, ego-centered network questions are the most
sophisticated. Respondents are asked to name a number of specified ties,
for example, the three best friends, and to report certain characteristics of
these ties. This information is then used to construct variables of interest
for the analyses; here, the relative share of majority members or coethnics
among the ties is the most frequently used indicator. Though relatively
time-expensive, the approach is in principle quite flexible, and can, for
example, target many different resources of the named persons so as to very
narrowly capture the concept of relevant social capital.
Within an explicit network design, however, sociometric questions ask
a respondent whether there are certain specified relations to each of the
other respondents. This then leads to a complete network specific for the
respective relation. Figure 1, for example, shows a friendship network for a
German classroom within the Children of Immigrants Longitudinal Survey
in Four European countries (CILS4EU, Kalter et al., 2015). It results from
the sociometric question:” Who are your best friends in class? “The colors
denote different ethnic backgrounds (yellow, majority without migration
background; green, Turkish background; and red, any other background).
The size of a dot represents the “strength” of a migration background (large,

Social Network Analysis in the Study of Ethnic Inequalities

5

8
14
9

2

7
4

18

11

15

6

3

10

13

16

17

Figure 1

1

5

12

Example of a classroom friendship network from the CILS4EU data.

born abroad; medium, at least one parent born abroad; and small, both
parents native-born). Arrows represent friendship nominations, pointing
from a nominator to the nominee.
Figure 1 shows, already intuitively, that the sociometric approach contains
very rich information about the social integration in the classroom. In general, it bears many well-known comparative advantages. In the following,
we stress four points that seem particularly important in the context of ethnic
minority research.
The first advantage is obviously that the information on the ties stems from
interviews with these persons themselves and is thus much more comprehensive and reliable. In integration research, the benefits of this fact already
start with the very concept of ethnicity itself. The evidence of a person’s
social integration as indicated by the share of minority members among the
nominated friends might differ according to the definition of ethnic minority
groups. In Figure 1, for example, students 6 and 14 would have equal value
(100%) when applying a wide concept. However, student 6 would have a
value lower than 14, if student 1, who belongs to the third generation, would
count as “majority” in a more narrow definition of ethnic minorities. In an

6

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

ego-centered approach, it seems almost impossible for a respondent to know
all the relevant details about his social ties needed to classify their ethnicity
correctly, for example, whether this person’s grandparents were born abroad,
or—as is common in the United Kingdom—someone subjectively identifies
with a certain group or not. Moving from the pure ethnicity of the social ties
to the social capital they include, the issues continue. This holds especially
when “softer” characteristics become theoretically relevant aspects, which is
very much the case within integration-related research questions—language
proficiency, cultural knowledge, or national identification of the social ties
being obvious examples.
The second, closely related point is that integration in general and social
integration in specific are not one-sided processes (Alba & Nee, 1997), but
heavily interactive. Information on how others, for example, majority classmates, treat a minority student at interest is obviously an important part of
the story. Looking back at Figure 1, the fact that student 6 is nominated as a
friend by a majority student (10) makes him, in a specific sense, somewhat
better integrated than student 14. Note that having information on the behavior and preferences of others also opens a door to measuring and studying
processes of discrimination in fresh new ways. This holds true all the more,
as the network approach allows us to examine very different kinds of relations between any two persons, among them so-called negative ties such as
disliking or bullying (Tolsma, van Deurzen, Stark, & Veenstra, 2013). Many
surveys still try to grasp discrimination by including questions on perceived
discrimination, despite long-standing doubts about the validity of these measures and thus the potential of drawing any causal conclusions (Blank et al.,
2004, p. 16ff).
The third advantage of the sociometric approach is that, casually speaking, it allows to study social integration at “higher orders”. Looking again at
our two exemplary students in Figure 1, student 6 is better socially integrated
than student 14 in the sense that he is befriended by students who themselves
have relatively many friends, among them majority members. Furthermore,
metaphorically speaking, overall he seems to be better “placed” within the
classroom than student 14. Social network analysis provides us now with a
rich repertoire of different measures that capture more precisely the general
position of an individual within a given network, for example, reachability,
actor centrality, or actor prestige (Wassermann & Faust, 1994). Such measures exist in many subvariants and can potentially be fine-tuned to more
adequately capture theoretically relevant aspects of social integration.
Fourth, and very importantly, the measures of social integration are not
restricted to being individual characteristics only. On the contrary, a major
attraction of network analysis is that it has developed many helpful measures describing characteristics of the network as a whole, such as density,

Social Network Analysis in the Study of Ethnic Inequalities

7

cliquishness, cohesiveness, and amount of clustering (Jackson, 2008, p. 20ff).
Many questions and discussions in integration research are related to concepts of social integration that are basically better understood as properties
of larger social units rather than of individual actors. A good example is the
general debate about the potential effects of ethnic diversity on social cohesion, stimulated by a meanwhile famous article by Robert Putnam (2007). The
empirical research on this topic typically employs individual-level survey
data using generalized trust, reported civic engagement, attitudes toward
the welfare states, or similar variables as indicators of social cohesion. Social
cohesion, however, is not only a psychological and attitudinal concept, but
importantly and foremost consists of relational aspects.
How ethnic diversity and social cohesion are empirically related thus
can and should also be addressed taking a bird’s eye view. For example,
in the network in Figure 1, the students in the classroom seem fairly well
connected. This can be specified more precisely in network analytical terms:
a very simple measure, for example, would be network density, which is
defined as the number of actual ties in the network divided by the number
of all ties possible. The value for the network in figure would be 0.18. One
can also attach numerical values to the ethnic diversity of the classroom.
Applying the meanwhile common Herfindahl-based index, the value for
the classroom in Figure 1 is 0.84. On the basis of this logic, Kalter and Kruse
(2015) show that in the representative sample of 184 classroom networks
in Germany within the CILS4EU study, there is no correlation between
ethnic diversity and network density. They also find similar results for the
representative classroom samples in England, the Netherlands, and Sweden.
Findings change only slightly upon application of more sophisticated indices
of social cohesion. The example thus shows that the network analytical
approach delivers a fresh, and in some respects more adequate empirical
view of important long-standing questions in integration research.
TOWARD THE MECHANISMS BEHIND SOCIAL INTEGRATION
If social integration is an important piece of the ethnic inequality puzzle, then
the factors accounting for it move into the focus of integration research. In
particular, thinking back to the above-mentioned idea of classical assimilation theory—it is important to analyze whether social integration is more
than just a mere consequence of the opportunity structure deriving from
structural integration, and, if so, why. Here, recently developed multivariate
models in network analysis meanwhile provide a rich toolbox of techniques.
One of the techniques that has already received considerable attention
for ethnic minority-related topics is Exponential Random Graph Models
(ERGM) or, synonymously, p* models (Robins, Pattison, Kalish, & Lusher,

8

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

2007). Basically, these apply a logistic regression approach to dyads of
network, thus allowing to assess which factors influence whether any two
persons within such a network have a certain relation or not. The explaining
factors can include characteristics of the individuals on either ends of a tie as
well as characteristics of the dyad itself, for example, homogeneity of both
individuals with respect to certain characteristics, whether they share other
ties, and more complex structural features.
This has accounted for much recent progress, especially in the study of
ethnic or racial homophily (Kruse, Smith, van Tubergen, & Maas, 2016;
Moody, 2001; Mouw & Entwisle, 2006; Stark, 2011; Wimmer & Lewis, 2010).
The finding is almost universal that friendships tend to be strongly ethnically
homogenous, which is often interpreted as an explicit preference of ethnic
group members: the unwillingness to integrate socially, on the part of the
minorities, or social discrimination, on the part of the majority. However,
the reasons for this can be very different and rather diverse and are hard
to disentangle using standard survey methods. First of all, it could simply
be a matter of opportunity structure. In this respect, the network analytical
approach in general and the ERGMs in particular allow a uniquely adequate
control, as the pool from which friends can be chosen at all is specified by the
complete network and, casually spoken, enters the denominator of the effect
estimation. Second, ethnic homogeneity can also result from preferences for
or the impact of characteristics that are only related to ethnicity, but should
not be confused with an explicit preference for ethnicity itself. For example,
students might prefer to befriend students from a similar social class background, as this means a similar economic situation of their families, allowing
them to engage in similar leisure activities. Similarly, additional and less
obvious structural arrangements might foster the ethnic homogeneity of
ties; students from the same ethnicity might for several reasons be more
likely to share different tracks within schools and extracurricular activities
or come from the same neighborhood. ERGM allows us to control for all this.
Finally, important built-in network mechanisms, sometimes called balancing mechanisms, can amplify the ethnic homogeneity of social ties, thus overestimating potentially underlying preference effects. Most basically, friendship choices and other sorts of social relations tend to be reciprocated. So
even if A chooses B because of an ethnic preference, the fact that B chooses
A might be due to simple reciprocity and itself not reveal any ethnic preference. Similarly, triads tend to closure, so if A and B have a relation and B and
C have a relation, it is very likely that A and C will also have a relation. It is
a strategic feature of ERGMs that they very conveniently and flexibly allow
to control for these kinds of “natural” network mechanisms when assessing
tendencies for ethnic friendship choices.

Social Network Analysis in the Study of Ethnic Inequalities

9

TOWARD DISENTANGLING THE COMPLEX RELATION BETWEEN
DIFFERENT ASPECTS OF INTEGRATION—SELECTION VERSUS
INFLUENCE
The network analytical approach also gives us a better methodological grip
on a key issue in integration research: the direction of causality between different aspects of integration. As sketched above, the literature suggests that
many mechanisms that consider certain aspects of social integration lead to
structural ethnic inequalities, such as the Granovetter type of social capital
effects on labor market outcomes. However, as also noted with reference to
classical assimilation theory, arguments in the other direction are at least
likewise plausible. Obviously, the answer to what is the right direction is
of major importance, not least for any integration-related policies and measures. Here, studying networks in longitudinal perspective offers some new
strategic tools of analysis that promise much better insights. They do so by
enabling us to empirically disentangle the two important general mechanisms of selection and influence.
Let us assume, for example, that we are dealing with two ethnic
groups—blues and yellows—and that blues on average do worse in school
than yellows. Let us also assume that, at a certain point in time (t), those
blues who perform better have relatively more yellows as friends. (If we
think of blues as representing an ethnic minority and yellows as representing
the majority this is empirically a quite frequent scenario.) This situation is
illustrated by some small networks in the right column (time t) in Figure 2,
where the size of a dot expresses school performance and lines represent
friendships. Basically, this relation could result from the fact that for the
yellows those blues who do better are more attractive choices as friends
and/or that the more successful blues are more likely to select yellows
as friends (because these are more likely good performers). This would
represent mechanisms of selection as illustrated by the transition from time
t − 1 to time t in Figure 2, where the focus is always on the blue in the middle.
It could also be, however, that those blues who are more strongly befriended
by yellows (for whatever reason) at time t − 1 profit from these ties and
become better performers at time t. This would correspond to the selection
mechanisms sketched in Figure 2.
Figure 2 thus shows that similar constellations at time t can arise from different starting points at time t − 1 via the different mechanisms of selection
versus influence, and that similar starting points at t − 1 can lead to different constellations at time t. Comparing networks over time thus allows to
distinguish selection effects from influence effects.
Meanwhile, sophisticated statistical techniques allow these ideas to be
applied to empirical network panel data. Tom Snijders and his workgroup

10

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Time t−1

Time t

on

ecti

Sel

Influ

enc

e

ion

ect

Sel

Infl

uen

ce

Figure 2 Illustration of mechanisms of selection and influence in small networks;
focus is on the blue dot in the middle.

developed stochastic actor-oriented models (SAOM) for network dynamics
(Snijders, van de Bunt, & Steglich, 2010) that permit uniquely strict tests.
Basically, these are agent-based simulation models assuming that actors
make choices about changing their social ties and their behavior. The decision
to form or cut specific ties and the decision for or against a specific change
in behavior are assumed to be dependent on a set of variables—pretty much
like in a regression model. Without going further into technical details of the
complex and assumption-rich estimation procedure, the important thing is
that this yields coefficient estimates and estimates for their standard errors
both for variables related to the network dynamics and, simultaneously, for
variables related to the behavior dynamics—among them coefficients that
directly capture the selection and influence mechanisms. Estimation can be
conducted with a software program called SIENA (Simulation Investigation
for Empirical Network Analysis), which is meanwhile integrated in the general
software package R and in this form also known under the name RSiena
(Ripley, Snijders, Boda, Vorce, & Preciado, 2015).
That SAOM techniques are meanwhile well established in many fields
is proved by the long list of published applications on the SIENA website

Social Network Analysis in the Study of Ethnic Inequalities

11

(www.stats.ox.ac.uk/∼snijders/siena/). A frequently researched topic,
for example, is the coevolution of adolescents’ friendship networks and
different forms of deviant behavior or delinquency, such as nicotine, drug,
and alcohol consumption, or carrying a weapon. With the rising shares of
ethnic minority students among the youth in almost all countries, variables
of ethnicity have occasionally also attracted interest, not only among the
controls. In the course of the vivid activities and progress, researchers are
increasingly aware of the potential of SAOM tools for addressing classical
questions of integration research, and some have already targeted these
more explicitly.
A recent example is the paper by Leszczensky, Stark, Flache, and
Munniksma (2015), which uses the data of the Arnhem School Study (TASS,
see Stark, 2011) to study the coevolution of friendship networks and host
country identification, thus disentangling the causal relationship between
aspects of social integration and cultural integration. Although there is an
empirical correlation between the strength of identification with the host
country and the number of native friends among adolescent immigrants,
the study does not support the hypothesis that influence mechanisms are
at work. On the selection side, there is also no evidence for the assumption
that immigrant youth who identify more strongly with the Netherlands are
more likely to prefer native-born Dutch friends. Interestingly, the analyses
show that, in turn, native-born Dutch significantly prefer immigrants who
identify more strongly with the Netherlands. The study is thus an especially
telling proof that in order to correctly understand the mechanisms behind
integration processes, it is important to consider relational aspects. While
the example focuses on the interplay between social and cultural aspects of
integration, it is easy to imagine similar interesting applications targeting
the relation between ethnic minorities’ social integration and structural
aspects, such as school performance and truancy.
NETWORK ANALYSIS, INTEGRATION RESEARCH, AND THE
PROGRAM OF ANALYTICAL SOCIOLOGY
In this essay, we have outlined the potential of social network analysis and
some new developments in that field for the study of ethnic inequalities. In
particular, we have exemplified how they can increase our understanding
of the precise role of social aspects in the complex process of integration.
Network analytical designs and techniques represent a promising complement to the survey data-based standard empirical research agenda; they provide particular adequate measures for diverse subaspects of the concept of
social integration and the recent development of sophisticated statistical network methods allows to tackle key open questions and issues of integration

12

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

research, such as disentangling the mechanisms of choice versus those of
opportunity structure or of selection versus influence.
The growing awareness and use of new network analytical tools in integration research is well in line with a more general theoretical movement and
orientation that has recently become known and more visible under the label
“analytical sociology” (Hedström & Bearman, 2009). Next to more epistemological messages emphasizing the general need for truly mechanism-based
explanations, a major impetus of this approach lies in the fact that—while
doubtless standing on the firm ground of methodological individualism—it
puts more attention and relative weight on the impact of social contexts and
social interactions rather than on all too deep details of action theories (Kalter
& Kroneberg, 2014).
Large-scale research survey data analysis has been a huge success story in
many fields, due not least to the fact that it can be guided by explicit and
elaborated theories of action (Kroneberg & Kalter, 2012). Human capital theory has always played a dominant role in migration and integration research
and, as briefly sketched at the beginning, is as a rule able to account for
much of what is going on in fields such as the labor market or education.
Despite all the progress, however, a major limitation of quantitative empirical ethnic inequality research has been that standard survey designs and
data sets force us to assume that actors behave independently in isolation
(Snijders & Steglich, 2015). Integration, however, is—like all interesting social
phenomena—a heavily interactive and highly complex dynamic process, in
which the actions of one actor immediately constitute an important context
and condition for the actions of the other, and so forth. The next generation of
empirical research must complement the traditional individual survey data
approach by designs and strategies of analysis that take these contextual and
interactional processes into explicit account. For many key open questions,
longitudinal network analytical designs seem to be among the most promising of these tools.
REFERENCES
Aguilera, M. B., & Massey, D. S. (2003). Social capital and the wages of Mexican
migrants: New hypotheses and tests. Social Forces, 82(2), 671–701.
Alba, R., & Nee, V. (1997). Rethinking assimilation theory for a new era of immigration. International Migration Review, 31(4), 826–874.
Blank, R. M., Dabady, M., & Citro, C. F. (Eds.) (2004). Measuring racial discrimination.
Washington, D.C.: National Academies Press.
Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal
of Sociology, 94(Supplement 95), 95–120.
Gordon, M. M. (1964). Assimilation in American life: The role of race, religion and national
origins. New York, NY: Oxford University Press.

Social Network Analysis in the Study of Ethnic Inequalities

13

Granovetter, M. S. (1973). The strength of weak ties. American Journal of Sociology,
8(6), 1360–1380.
Heath, A., & Brinbaum, Y. (Eds.) (2014). Unequal attainments. Ethnic educational
inequalities in ten western countries. Oxford, England: British Academy; Oxford University Press.
Heath, A., & Cheun, S. Y. (Eds.) (2007). Unequal chances. Ethnic minorities in western
labour market. Oxford, England: British Academy; Oxford University Press.
Heath, A. F., Rothon, C., & Kilpi, E. (2008). The second generation in Western Europe:
Education, unemployment, and occupational attainment. Annual Review of Sociology, 34, 211–235.
Hedström, P., & Bearman, P. (Eds.) (2009). The Oxford handbook of analytical sociology.
New York, NY: Oxford University Press.
Jackson, M. O. (2008). Social and economic networks. Princeton, NJ: Princeton University Press.
Kalter, F. (2011). The second generation in the German labor market. In R. Alba & M.
C. Waters (Eds.), The next generation: Immigrant youth in a comparative perspective.
New York: NYU Press.
Kalter, F., Heath, A., Hewstone, M., Jonsson, J. O., Kalmijn, M., Kogan, I., & van
Tubergen, F. (2015). Children of immigrants longitudinal survey in four European
countries (CILS4EU). GESIS Data Archive, Cologne. ZA5353 Data file Version
2.1.0, doi: 10.4232/cils4eu.5353.2.1.0
Kalter, F., & Kroneberg, C. (2014). Between mechanism talk and mechanism cult:
New emphases in explanatory sociology and empirical research. In J. Friedrichs
& A. Nonnenmacher (Eds.), Kölner Zeitschrift für Soziologie und Sozialpsychologie,
Special Issue 54: Social Contexts and Social Mechanisms (pp. 91–115). Wiesbaden, Germany: Springer VS.
Kalter, F., & Kruse, H. (2015). Ethnic diversity, homophily, and network cohesion
in European classrooms. In R. Koopmans, B. Lancee & M. Schaeffer (Eds.), Social
cohesion and immigration in Europe and North America: mechanisms, conditions, and
causality. London, England: Routledge.
Kanas, A., Chiswick, B. R., van der Lippe, T., & van Tubergen, F. (2012). Social contacts and the economic performance of immigrants: A panel study of immigrants
in Germany. International Migration Review, 46(3), 680–709.
Kroneberg, C., & Kalter, F. (2012). Rational choice theory and empirical research.
Methodological and theoretical contributions in Europe. Annual Review of Sociology, 38, 73–92.
Kruse, H., Smith, S., van Tubergen, F., & Maas, I. (2016). From neighbours to school
friends? How adolescents’ place of residence relates to same-ethnic school friendships. Social Networks, 44(1), 130–142.
Lancee, B. (2010). The economic returns of immigrants’ bonding and bridging social
capital: The case of the Netherlands. International Migration Review, 44(1), 202–226.
Lancee, B. (2012). The economic returns of bonding and bridging social capital for
immigrant men in Germany. Ethnic and Racial Studies, 35(4), 664–683.

14

EMERGING TRENDS IN THE SOCIAL AND BEHAVIORAL SCIENCES

Leszczensky, L., Stark, T. H., Flache, A., & Munniksma, A. (2015). Disentangling
the relationship between young immigrants’ host country identification and their
friendships with natives. Social Networks, 44, 179–189.
Moody, J. (2001). Race, school integration, and friendship segregation in America.
American Journal of Sociology, 107(3), 679–716.
Mouw, T., & Entwisle, B. (2006). Residential segregation and interracial friendship
in schools. American Journal of Sociology, 112(2), 394–441.
Portes, A. (1995). Economic sociology and the sociology of immigration: A conceptual overview. In A. Portes (Ed.), The economic sociology of immigration (Vol. 29, pp.
1–41). New York, NY: Russell Sage Foundation.
Portes, A., & MacLeod, D. (1999). Educating the second generation: Determinants of
academic achievement among children of immigrants in the United States. Journal
of Ethnic and Migration Studies, 25(3), 373–396.
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.
Ripley, R.M., Snijders, T.A.B., Boda, Z., Vorce, A., & Preciado, P. (2015). Manual for RSiena. Retrieved from http://www.stats.ox.ac.uk/∼snijders/siena/
RSiena_Manual.pdf
Robins, G., Pattison, P., Kalish, Y., & Lusher, D. (2007). An introduction to exponential
random graph (p*) models for social networks. Social Networks, 29, 173–191.
Snijders, T. A. B., & Steglich, C. E. G. (2015). Representing micro–macro linkages by
sector-based dynamic network models. Sociological Methods & Research, 44, 22–271.
Snijders, T. A. B., van de Bunt, G. G., & Steglich, C. E. G. (2010). Introduction to
stochastic actor-based models for network dynamics. Social Networks, 32, 44–60.
Stark, T. H. (2011). Integration in schools. A process perspective on students’ interethnic
attitudes and interpersonal relationships (ICS dissertation). Groningen.
Tolsma, J., van Deurzen, J., Stark, T. H., & Veenstra, R. (2013). Who is bullying whom
in ethnically diverse primary schools? Exploring links between bullying, ethnicity,
and ethnic diversity in Dutch primary schools. Social Networks, 36, 51–61.
Wassermann, S., & Faust, K. (1994). Social network analysis: Methods and applications
(structural analysis in the social sciences (Vol. 8). Cambridge, England: Cambridge
University Press.
Wimmer, A., & Lewis, K. (2010). Beyond and below racial homophily: ERG models
of a friendship network documented on Facebook. American Journal of Sociology,
116, 583–642.

FRANK KALTER SHORT BIOGRAPHY
Frank Kalter is a professor of sociology at the School of Social Sciences at the
University of Mannheim, Germany. Currently, he also serves as the director
of the Mannheim Centre for European Social Research (MZES). He is a fellow
of the European Academy of Sociology (EAS) and has been its president from
2011 to 2015. His major research interests include migration, the integration of
immigrants and their children, and the formal modeling of social processes.

Social Network Analysis in the Study of Ethnic Inequalities

15

RELATED ESSAYS
Misinformation and How to Correct It (Psychology), John Cook et al.
Enduring Effects of Education (Sociology), Matthew Curry and Jennie E.
Brand
Cognitive Processes Involved in Stereotyping (Psychology), Susan T. Fiske
and Cydney H. Dupree
Controlling the Influence of Stereotypes on One’s Thoughts (Psychology),
Patrick S. Forscher and Patricia G. Devine
Culture and Social Networks (Sociology), Jan A. Fuhse
Migrant Networks (Sociology), Filiz Garip and Asad L. Asad
Ethnic Enclaves (Sociology), Steven J. Gold
Immigrant Children and the Transition to Adulthood (Sociology), Roberto G.
Gonzales and Benjamin J. Roth
Racial Disenfranchisement (Political Science), Vincent L. Hutchings and
Davin L. Phoenix
Emerging Trends in Social Network Analysis of Terrorism and Counterterrorism (Sociology), David Knoke
Immigration and the Changing Status of Asian Americans (Sociology),
Jennifer Lee
Immigrant Sociocultural Adaptation, Identification, and Belonging (Sociology), Sarah J. Mahler
Political Inequality (Sociology), Jeff Manza
Immigrant Health Paradox (Sociology), Kyriakos S. Markides and Sunshine
Rote
Politics of Immigration Policy (Political Science), Jeannette Money
The Role of School-Related Peers and Social Networks in Human Development (Political Science), Chandra Muller
Health and Social Inequality (Sociology), Bernice A. Pescosolido
Latinos and the Color Line (Sociology), Clara E. Rodríguez et al.
Digital Methods for Web Research (Methods), Richard Rogers
Stereotype Threat (Psychology), Toni Schmader and William M. Hall
Assimilation and Its Discontents (Sociology), Min Zhou