Residential segregation and ‘ethnic flight’ vs. ‘ethnic avoidance’ in Sweden Tim S. Müller, Humboldt University Berlin and Linköping University Thomas U. Grund, University College Dublin and Linköping University Johan Koskinen, University of Manchester and Linköping University Tim Müller (corresponding author) Humboldt University Berlin Berlin Institute for Integration and Migration Research (BIM) Unter den Linden 6 10099 Berlin, Germany [email protected]Thomas Grund University College Dublin School Of Sociology Newman Building Belfield Dublin 4, Ireland [email protected]Johan Koskinen Social Statistics Discipline Area School of Social Sciences Humanities Bridgeford Street University of Manchester MANCHESTER M13 9PL, United Kingdom [email protected]Word count: 8224 First submitted on 12/08/2015 First revision submitted on 30/10/2016 Second revision submitted on 28/04/2017 Third revision submitted on 29/11/2017 Final revision submitted on 23/03/2018
40
Embed
Residential segregation and ‘ethnic flight’ vs. ‘ethnic ... · Web viewJohan Koskinen. Social Statistics Discipline Area. School of Social SciencesHumanities Bridgeford StreetUniversity
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Residential segregation and ‘ethnic flight’ vs. ‘ethnic avoidance’ in SwedenTim S. Müller, Humboldt University Berlin and Linköping UniversityThomas U. Grund, University College Dublin and Linköping UniversityJohan Koskinen, University of Manchester and Linköping University
Tim Müller (corresponding author)Humboldt University BerlinBerlin Institute for Integration and Migration Research (BIM)Unter den Linden 610099 Berlin, [email protected]
Thomas GrundUniversity College DublinSchool Of SociologyNewman BuildingBelfieldDublin 4, [email protected]
Johan KoskinenSocial Statistics Discipline AreaSchool of Social SciencesHumanities Bridgeford StreetUniversity of ManchesterMANCHESTERM13 9PL, United [email protected]
Word count: 8224First submitted on 12/08/2015First revision submitted on 30/10/2016Second revision submitted on 28/04/2017Third revision submitted on 29/11/2017Final revision submitted on 23/03/2018
Funding: This research has received funding from the European Research Council under the European Union's Seventh Framework Programme (FP7/2007-2013) / ERC grant agreement no 324233, Riksbankens Jubileumsfond (DNR M12-0301:1), and the Swedish Research Council (DNR 445-2013-7681) and (DNR 340-2013-5460).
that could not be discerned from the individual level. Therefore, we apply several
different modelling strategies to explain individual moving-out and moving-in
behaviour. Firstly, we use a two-step hierarchical estimation technique, suggested by
Achen (2005) and Lewis and Linzer (2005), which allows controlling for individual
and neighbourhood characteristics. Secondly, we use multilevel logistic and
multinomial logistic regression models to allow for the comparison of more fine-
grained categories of immigrant background and their different reactions towards
changes in the ethnic neighbourhood composition, as previous research hints at
more differentiated inter-ethnic preferences (cf. Crowder, 2000; Pais, South, and
Crowder, 2009).
Modelling individual moves
For the two-step hierarchical regression analysis first individual-level logistic
regression models predicting the decision to leave (or stay) are performed for each
neighbourhood and each year. We control for a broad range of possible individual
confounding variablesiv and include a variable that distinguishes between native
Swedish vs. immigrant background of a person. The logit coefficient for this covariate
becomes the dependent variable in the second step of the analysis, which performs
neighbourhood-level regressions to account for variation in coefficients between
neighbourhoods and over time.v We apply Fixed Effects WLS panel regressions with
the following explanatory variables on the neighbourhood level: %Immigrant, logged
population size of the neighbourhood, and the median disposable household income
of each neighbourhood.
TABLE 2 ABOUT HERE
Moving-out results
Figure 4 displays the bivariate relationship between the logarithmic odds for leaving
a neighbourhood for immigrant vs native from the first step of the two-step
hierarchical estimation (individual-level). Logistic regression coefficients for each
neighbourhood are plotted against the share of residents with immigrant background
in each neighbourhood. Positive values indicate that immigrants are more likely to
leave, while negative values indicate that Swedish natives are more likely to leave a
neighbourhood. There exists a small effect of neighbourhood ethnic composition on
individuals’ chances to leave, even after controlling for individual-level
characteristics. Native Swedes are slightly more likely to leave a neighbourhood
where many immigrants live compared to immigrants. However, for most
neighbourhoods the effect is close to 0.
FIGURE 4 ABOUT HERE
Looking at the distribution of coefficients (see Figure 4), we cannot discern
discontinuities which would indicate tipping points. Hence, our findings do not align
with Aldén et al. (2015), who found a sudden increase in moving-out behaviour once
the share of immigrants reached 4.5 to 9% in a neighbourhood.
Table 2 shows results from the second step (neighbourhood-level) of the two-step
hierarchical estimation for various model specifications. Results indicate that
immigrants are less likely to leave a neighbourhood compared to native Swedes
when the share of immigrants goes up. Similarly, immigrants are less likely to leave
than native Swedes when the neighbourhood median income increases.
Relaxing the dichotomous distinction between immigrants and native Swedes, we
ran multilevel logistic regression analyses with similar controls for three ethnic
groups: native Swedish, European Union (EU) and non-European Union (non-EU)
background. These analyses included interaction terms between the individual ethnic
group and the neighbourhood-level ethnic composition.vi To address the possibility
that individuals might base their moving out decision not on the static neighbourhood
composition but on the change in composition between t-2 and t-1, we also
calculated models with the respective difference in composition.
FIGURE 5 ABOUT HERE
Figure 5 shows the odds-ratio (leaving vs. staying) for Swedish natives (left),
European Union immigrants (middle) and non-European Union immigrants (right)
when either the share of EU or non-EU immigrants increases at the neighbourhood
level for each year. Coefficients represent the effect strength of an increase in group
size by one standard deviation in a given year (increase in the difference of group
size between t-1 and t0) compared within each group.vii For the Swedish native
population, we find significant ethnic flight effects with regard to the non-EU
population, but only for the first three years under examination (upper panel, left plot
in Figure 5). For the next few years, the effect is essentially zero and even becomes
negative in the years 1998-2002. Increases in the EU population seem to lead to a
slightly higher chance of leaving from 1995 onwards. There is some evidence for
ethnic flight behaviour for the Swedish population, but the pattern is not clear over
time. We find similar results for the moving-out chances of EU migrants (upper
panel, middle plot in Figure 5). There are no more flight effects from 1994 onwards.
For non-EU immigrants (upper panel, right plot in Figure 5), there is also no time-
consistent moving-out pattern.
While the pattern for changes in composition between t0 and t-1 is slightly different
(lower panel in Figure 5), the overall picture does not change substantially. The
compositional change in EU or non-EU immigrants in a neighbourhood does
basically not affect the moving-out chances for a Swedish native individual (lower
panel, first plot in Figure 5). There are also only small and inconsistent effects on the
moving-out chances for the other groups. Overall, moving-out patterns do not seem
to point to any significant “ethnic flight” behaviour on part of any of the groups.
FIGURE 6 ABOUT HERE
Moving-in results
To test, whether Swedes avoid moving into neighbourhoods where many immigrants
live (“ethnic avoidance”), we use a similar design as before. But this time the
population of interest comprises those individuals who move into a new
neighbourhood.viii The dependent variable in the first step regression is “immigrant
vs. native”. We control for the same set of individual-level variables as before.
Figure 6 shows the log-odds for being an immigrant vs. native Swedish among the
population of individuals who move towards a new neighbourhood against the
proportion of immigrants living in these new neighbourhoods. Results suggest
“ethnic avoidance” behaviour; in neighbourhoods with small shares of immigrants,
natives have higher chances to be among the in-moving population compared to
immigrants. In neighbourhoods where many immigrants live, immigrants are more
likely to be among the in-movers. The second step neighbourhood-level regressions
confirm these findings controlling for all other individual and neighbourhood
characteristics (Table 2).
FIGURE 7 ABOUT HERE
Relaxing our distinction between natives and immigrants, the additional multinomial
regressionsix explore if it is undifferentiated avoidance behaviour (native vs.
immigrant) or whether different preferences between more fine-grained categories
(native Swedish, EU, non-EU) exist. On the macro-level we controlled for median
neighbourhood income and the share of EU- and non-EU residents in the destination
neighbourhoods in the year prior to the move. The results in Figure 7 present
comparisons with different baseline categories, which can be easily derived from the
original model results. In the left plot (upper panel, Figure 7) we find the results of a
native Swede being an in-mover compared to a person with non-EU background.
The different symbols refer to the effects that a one standard deviation increase in
either the EU or non-EU population have on moving into the neighbourhood. The
pattern is clear and consistent. Most of the time, the chances are halved for a
Swedish native compared to a non-EU immigrant to be found among the in-mover
population as the share of non-EU residents increases by one standard deviation.
However, Swedes have about the same chance as non-EU immigrants to move into
neighbourhoods with an increase in EU migrant population by one standard
deviation.
The second plot (upper panel) of Figure 7 shows that EU-immigrants consistently
have a higher chance to move into neighbourhoods with a higher share of either EU-
or non-EU population than native Swedes. And lastly the third plot (upper panel) of
Figure 7 shows that as the share of non-EU residents in a destination neighbourhood
increases by one standard deviation, the odds are almost doubled for a non-EU
individual to be an in-mover compared to Swedish natives. But native Swedes and
non-EU immigrants have about the same chance to move into a neighbourhood with
a higher share of EU-residents. The results for the compositional changes between t-
2 and t-1 (lower panel) generally confirm this picture. Swedish natives (left plot, lower
panel) avoid moving into neighbourhoods that experienced an increase of either EU
or non-EU populations in the previous period. EU and non-EU immigrants are more
likely to move into neighbourhoods that experienced an increase in EU or non-EU
populations respectively (middle and right plots, lower panel).
Conclusion
This article aims to investigate the origins of segregation in Sweden (see also
Bråmå, 2006, 2008). Based on the most common explanations for selective in- and
out-movement patterns, ethnic preferences (Schelling, 1971; Emerson et al., 2001),
discrimination (Massey and Denton, 1993; Zubrinsky and Bobo, 1996), socio-
economic differences (Crowder, 2000) and previous research we investigate two
mechanisms in detail: 1) “ethnic flight”, which refers to the selective out-movement of
natives from neighbourhoods where many immigrants live, and 2) “ethnic
avoidance”, which refers to selective in-movement of natives to neighbourhoods
where only few immigrants live. We apply a two-pronged strategy. First, we
conceptualise the flows of movement of Swedes and immigrants between Stockholm
neighbourhoods between 1990 and 2003 as a network and apply exponential
random graph models. This macro-level approach allows us to account for hierarchy
between neighbourhoods as well as spatial dependence of moves, which go
unnoticed at the individual level. While residential moves between neighbourhoods
aggregate into flows of stocks from one location to another, aggregate flows reveal
repeated structural patterns of exchange, much like roads may be considered
aggregates of traffic and ant paths are emergent highways. The most travelled paths
also are indicative of systemic constraints – not everyone can live in the same
neighbourhood. Second, we complement these analyses with micro-level analyses
at the individual level. These analyses cannot account for complex
interdependencies between moves, but they allow for the inclusion of characteristics
of individuals and neighbourhoods. It is clear that both levels of analysis are
important but a new modelling framework to combine both would be required.x
On the macro-level, we find clear evidence for “ethnic avoidance”. Swedes are more
likely to move towards neighbourhoods where fewer immigrants live. Surprisingly,
this effect is even more pronounced when controlling for socio-economic conditions
at the neighbourhood level. There is no evidence for “ethnic flight”. On the micro-
level, we also find support for “ethnic avoidance”. Individuals who move to a new
neighbourhood are more likely to be immigrants than Swedes when the share of
immigrants is high in the destination neighbourhood. Looking at the total population,
Swedes are slightly more likely to leave neighbourhoods where many immigrants
live. Hence, we only find scant evidence for “ethnic flight” at the individual-level. To
summarise, our findings suggest that “ethnic avoidance” and not “ethnic flight” is the
main driver behind segregation in Sweden.
The application of the network approach is adding to the existing literature in the field
by explicitly taking into account (1) that a popularity hierarchy between
neighbourhoods might exist, which structures movement decisions, but which is
usually not considered in models of individual movement decisions. Such a hierarchy
might not accurately be reflected by observable variables, but it can be inferred from
the topography of the network of flows; (2) that movements in one part of the city
might structure the alternatives of movers in other parts of the city (similar to
Harrison White’s, 1970, argument of vacation chains); (3) that movements are
strongly dependent on spatial proximity. Generally, we find that ignoring these
effects leads to a biased estimation of the other effects.
How do our results line up with previous research? First, the effects of “ethnic flight”
are very small or hardly detectable in the analysis of movement flows. This is not an
unusual finding. Even in the United States, where segregation is more pronounced,
the observed effect for “ethnic flight” is very small (Crowder, 2000; Quilian, 2002;
Crowder et al. 2011). The results of our individual-level analyses show that “ethnic
flight” is detectable in Stockholm, but the effects are far too small to exert a
meaningful influence on the segregation process. This also holds true if one looks at
compositional differences rather than static neighbourhood compositions. The effects
of selective in-movement and “ethnic avoidance” seem much more important. This is
also in line with previous research (Andersson, 2013; Quilian, 2002; Hedman and
Ham, 2011; Bråmå, 2006; Simpson and Finney 2009). More recently, Aldén et al.
(2015) suggested that “ethnic flight” and not “ethnic avoidance” drives segregation in
Sweden after 2000. This inconsistency with our results could be due to different time
periods or our focus on Stockholm municipality, which largely omits urban/rural
differences. Most remarkably, our analyses indicate that the avoidance effects
increase after controlling for the income in neighbourhoods. While more research is
certainly needed, previous studies also show that ethnic or racial preferences persist
after taking socio-economic conditions into account (Crowder, 2000; Crowder et al.,
2011; Emerson et al., 2001). In consequence, the observed levels of segregation
would not be reduced by policies that strictly aim at the reduction of poverty or
neighbourhood distress (Andersson and Bråmå, 2004; Andersson, 2006). It would
need a change in “ethnic preferences” to reduce ethnic avoidance behaviour. A
further explanation for large differences in moving-in patterns might be found by
taking into account the properties of the Swedish housing system, which allocates
rental housing according to waiting time and therefore might put newcomers to the
Stockholm housing market at a disadvantage (cf. Özüekren and van Kempen 2003;
Andersen et al. 2013). The market might then be divided between native Swedish
tenants with longer queue waiting times, which can get easier access to the inner
central neighbourhoods (also by exercising their option to buy apartments) and
newcomers (many of them immigrants) that due to their shorter waiting times might
be confined to the more peripheral neighbourhoods, resulting in the moving patterns
that we have observed in this study.
ReferencesAchen, C. H. (2005). Two-Step Hierarchical Estimation: Beyond Regression Analysis, Political Analysis, 13(4), 447-456.
Ahmed, A., Andersson, L. and Hammarstedt, M. (2010) Can ethnic discrimination in the housing market be reduced by increasing the information about the applicants? Land Economics, 86(1), 79–90.
Ahmed, A. and Hammarstedt, M. (2008) Discrimination in the Rental Housing Market: A Field Experiment on the Internet. Journal of Urban Economics, 64(2), 362-372.
Aldén, L., Hammarstedt, M. and Neuman, E. (2015) Ethnic Segregation. Tippingf Behavior, and Native Residential Mobility, International Migration Review, 49(1), 36-69.
Andersen, H. S., Turner, L. M., & Søholt, S. (2013). The special importance of housing policy for ethnic minorities: evidence from a comparison of four Nordic countries. International Journal of Housing Policy, 13(1), 20-44.
Andersson, R. (2013) Reproducing and reshaping ethnic residential segregation in Stockholm: the role of selective migration moves. Geografiska Annaler: Series B, Human Geography, 95(2), 163-187
Andersson, R. and Bråmå, Å. (2004). Selective migration in Swedish distressed neighbourhoods: can area-based urban policies counteract segregation processes? Housing Studies, 19(4), 517-539.
Andersson, R. (2006). ‘Breaking Segregation’—Rhetorical Construct or Effective Policy? The Case of the Metropolitan Development Initiative in Sweden, Urban Studies, 43(4), 787–799.
Bråmå, Å. (2006). ’White Flight’? The Production and Reproduction of Immigrant Concentration Areas in Swedish Cities, 1990-2000, Urban Studies, 43(7), 1127-1146.
Bråmå, Å. (2008). Dynamics of ethnic residential segregation in Göteborg, Sweden, 1995–2000. Population, Space and Place, 14(2), 101-117.
Bruch, E. and Mare, R. D. (2006). Neighborhood Choice and Neighborhood Change, American Journal of Sociology, 112(3), 667-709.
Bygren, M. and Kumlin, J. (2005). Mechanisms of organizational sex segregation: organizational characteristics and the sex of newly recruited employees, Work and Occupations, 32, 39-65.
Card, D., Mas, A., and Rothstein, J. (2008). Tipping and the Dynamics of Segregation. The Quarterly Journal of Economics, 123(1), 177-218.
Crowder, K. (2000). The Racial Context of White Mobility: An Individual-Level Assessment of the White Flight Hypothesis, Social Science Research, 29, 223–257.
Crowder, K., Hall, M. and Tolnay, S. E. (2011). Neighborhood Immigration and Native Out-Migration, American Sociological Review, 76(1), 25-47.
Crowder, K. and South, S. J. (2008). Spatial Dynamics of White Flight: The Effects of Local and Extralocal Racial Conditions on Neighborhood Outmigration, American Sociological Review, 73(5), 792-812.
Daraganova, G., Pattison, P., Koskinen, J., Mitchell, B., Bill, A., Watts, M., Baum, S. (2012). Networks and geography: modelling community network structures as the outcome of both spatial and network processes, Social Networks, 34(1), 6-17.
De Benedictis, L., Tajoli, L. (2011) The world trade network. World Econ 34:1417–1454
Ellen, I. G. (2000). Sharing America’s neighborhoods. Cambridge, MA: Harvard University Press.
Emerson, M. O., Chai, K. J. and Yancey, G. (2001). Does race matter in residential segregation? Exploring the preferences of White Americans, American Sociological Review, 66(6), 922–935.
Fagiolo, G., Mastrorillo, M. (2013). International migration network: Topology and modeling, Physical Review E, 88, 012812.
Galster, G. (1988). Residential segregation in American cities: A contrary Review, Population Research and Policy Review, 7(2), 93-112.
Galster, G. (1990). White flight from racially integrated neighborhoods in the 1970s: The Cleveland experience, Urban Studies, 27, 385-399.
Goldstein, H., & Noden, P. (2003). Modelling social segregation. Oxford Review of Education, 29(2), 225-237.
Groshen, E. L. (1991). The structure of the female/male wage differential. Is it who you are, what you do, or where you work? Journal of Human Resources, 26(3), 457-472.
Hedman, Lina and Ham, M. (2011). Neighbourhood Choice and neighbourhood reproduction, Environment and Planning A, 43, 1381-1399.
Holland, P.W. and Leinhardt, S.(1970). A Method for Detecting Structure in Sociometric Data. American Journal of Sociology, 76(3):492-513
Hunter, D. R.,Goodreau, S. M., and Handcock, M. S. (2008). Goodness of fit of social network models. Journal of the American Statistical Association, 103, 248–258.
Koskinen, J., and Lomi, A. (2013). The Local Structure of Globalization: The Network Dynamics of Foreign Direct Investments in the International Electricity Industry. Journal of Statistical Physics. Vol. 151, (3), 523-548.
Koskinen, J., Mueller, T., Grund, T. (in press). A dynamic discrete-choice model for movement flows. In Perna, C., Pratesi, M. & Ruiz-Gazen, A. (eds.), Studies in Theoretical and Applied Statistics. Springer
Leckie, G., & Goldstein, H. (2015). A multilevel modelling approach to measuring changing patterns of ethnic composition and segregation among London secondary schools, 2001–2010. Journal of the Royal Statistical Society: Series A (Statistics in Society), 178(2), 405-424.
Lewis, J. B. and Linzer, D. A. (2005). Estimating Regression Models in Which the Dependent Variable Is Based on Estimates, Political Analysis, 13, 345-364.
Lusher, D., Koskinen, J. and Robins, G. (2013) (Eds.). Exponential Random Graph Models for Social Networks: Theory, Methods and Applications, Cambridge University Press: New York.
Massey, D. S. and Denton, N. A (1993). American Apartheid. Segregation and the Making of the Underclass. Cambridge, MA: Harvard University Press.
McPherson, M., Smith-Lovin, L., and Cook, J. M. (2001). Birds of a feather: Homophily in social networks, Annual Review of Sociology, 27, 415–444.
Merton,R.K. (1968) The Matthew effect in science. Science, 159, 56–63.
Özüekren, A. S., & Van Kempen, R. (2002). Housing careers of minority ethnic groups: Experiences, explanations and prospects. Housing studies, 17(3), 365-379.
Ondrich, J., Stricker, A. and Yinger, J. (1999). Do Landlords Discriminate? The Incidence and Causes of Racial Discrimination in Rental Housing Markets, Journal of Housing Economics, 8(3), 185-220.
Pais, J. F., South, S. J. and Crowder, K. (2009). White Flight Revisited: A Multiethnic Perspective on Neighborhood Out-Migration, Population and Residential Policy Review, 28, 321-346.
Quilian, L. (2002). Why Is Black–White Residential Segregation So Persistent? Evidence on Three Theories from Migration Data, Social Science Research, 31, 197-229.
Robins, G.L., Elliott, P., & Pattison, P.E. (2001). Network models for social selection processes, Social networks, 23, 1–30.
Robins, G., Pattison, P., & Wang, P. (2009). Closure, connectivity and degree distributions: Exponential random graph (p*) models for directed social networks. Social Networks, 31(2), 105-117.
Robins, G. L., Lusher, D., 2013. Illustrations: Simulation, Estimation, and Goodness of Fit. In: Lusher, D., Koskinen, J. H., Robins, G. L. (Eds.), Exponential Random Graph Models for Social Networks: Theory, Methods, and Applications. Cambridge University Press, Cambridge, UK, pp. 167–185.
Ross, S. L. and Austin Turner, M. (2005). Housing discrimination in metropolitan America: Explaining changes between 1989 and 2000, Social Problems, 52(2), 152-180.
Sampson, R. J. and Sharkey, P. (2008). Neighborhood Selection and the Social Reproduction of Concentrated Racial Inequality, Demography, 45(1), 1- 29.
Schelling, T. C. (1971). Dynamic models of segregation, Journal of Mathematical Sociology, 1(2), 143-186.
Schweinberger, M., Krivitsky, P. N., and Butts, C. T. (2017). Foundations of finite-, super-, and infinite-population random graph inference. https://arxiv.org/abs/1707.04800.
Sharkey, P., & Elwert, F. (2011). The legacy of disadvantage: Multigenerational neighborhood effects on cognitive ability. American Journal of Sociology, 116(6), 1934-81.
Simpson, L. and Finney, N. (2009). Spatial patterns of internal migration: evidence for ethnic groups in Britain, Population, Space and Place 15 (1), 37– 56.
Slater, P. B. (2008). Hubs and clusters in the evolving us internal migration network. arXiv preprint arXiv:0809.2768.
Snijders, T.A.B. (2001). The statistical evaluation of social network dynamics. Socio- logical Methodology, 361–395, Vol 31 31.
Snijders, T.A.B., and van Duijn, M.A.J. (2002). Conditional Maximum Likelihood Estimation under Various Specifications of Exponential Random Graph Models. Iin Hagberg, J. (ed.), Contributions to Social Network Analysis, Information Theory, and Other Topics in Statistics; A Festschrift in honour of Ove Frank. University of Stockholm, Department of Statistics, 117-134.
Snijders, T. A. B., Pattison, P., Robins, G. and Handcock. M. (2006). New specifications for exponential random graph models, Sociological Methodology, 36(1), 99-153.
South, S. J. and Crowder, K. (1998) Leaving the 'Hood: Residential Mobility between Black, White, and Integrated Neighborhoods, American Sociological Review, 63(1), 17-26.
Van de Rijt, A., Siegel, D. and Macy, M. (2009). Neighborhood Chance and Neighborhood Change: A Comment on Bruch and Mare, American Journal of Sociology, 114(4), 1166-1180.
Wasserman, S. and Pattison, P. (1996). Logit Models and Logistic Regressions for Social Networks. An Introduction to Markov Graphs and p*, Psychometrika, 61(3), 401-42.
White, H. (1970). Chains of Opportunity. System Models of Mobility in Organizations. Cambridge, MA: Harvard University Press.
Wilson, W.J. 1987. The Truly Disadvantaged: Essays on Inner City Woes and Public Policy. Chicago: University of Chicago Press.
Wimmer, A. and Lewis, K. (2010). Beyond and Below Racial Homophily: ERG Models of a Friendship Network Documented on Facebook, American Journal of Sociology, 116(2), 583-642.
Zubrinsky C. and Bobo, L. (1996). Prismatic Metropolis: Race and Residential Segregation in the City of the Angels, Social Science Research, 25: 335-374.
i Average disposable household income (from all income sources after taxes) is calculated over the complete
period in our data and standardized by household size. We are using the component that is calculated for
each household member and take the arithmetic mean for each neighbourhood. This measure serves as a
proxy for neighbourhood socio-economic conditions, but could also reflect rent prices
ii We also ran additional models, which omitted the structural network effects for systemic dependencies in
the network of movement flows (two-paths, transitive and cyclical triads, which are used to model the
underlying popularity hierarchy among neighbourhoods). (See Table A2 in the online supplement.)
Noteworthy differences emerge, when these effects are not explicitly taken into account: (1) for the immigrant
population the hierarchy effect is erroneously attributed to homophily on income (Model 6 vs. Model 8); (2)
the effect of nbhd. population size (i.e. vacancies) is underestimated; (3) the results point towards a (non-
significant) ethnic flight effect among the native population; (4) the effect of nbhd. income as a negative
predictor of outflows is attenuated while income in receiving nbhds. as a negative predictor of inflow is
inflated; (5) generally, the effects of the ethnic nbhd. composition as a characteristic of receiving nbhds. are
attenuated for native and immigrant movers. These differences are also reflected in the goodness-of-fit. (For
a discussion of GOFs see: Hunter et al., 2008.) The model without hierarchy produces networks that are not
as clustered as the observed network and where the ties of the immigrant population are much more evenly
spread out across neighbourhoods than they actually are. Furthermore, where M6 neither captures the
degree distribution nor the clustering coefficients, M8 replicates all of these.
iii In general, the fit for the models including the additional structural effects was better in comparison to the
models that omitted them and the inclusion of nbhd. income improved the model fit compared to the models
omitting nbhd. Income, following the criteria of Robins and Lusher (2013:184-185). A comparison of model fit
by more conventional global measures (e.g. AIC) is difficult and currently subject to debate (Schweinberger
et al., 2017).
iv The nesting is individuals in neighborhoods. We control for a range of socio-demographic characteristics
that have been found to be important in this context (South and Crowder, 1998): age, disposable household
income adjusted for household size, sex, number of children below 18, marital status and immigrant
background. The online supplement (sections A4 and A5) contains further information on the models and
descriptive statistics.
v We are using Achen’s (2005) and Lewis and Linzer’s (2005) approach to assure consistency and efficiency
from the second step regression by weighting for the sampling error of the first stage. Further panel model
specifications with very similar results are presented in Tables A3.1 and A3.2 in the appendix. They also
include models with time-lagged variables.vi This means the effect sizes of the compositional variables are plotted. For Swedish natives the main effect is plotted, for the other groups the interaction terms of the compositional variable with the group variable (EU/non-EU) are plotted. On the neighbourhood-level we accounted for the share of EU and non-EU immigrants in the year of moving, the neighbourhood median HH income and the population size. Individual-level controls are the same as in the two-step procedure (see online supplement A4 and A5).
vii E.g. a unit increase in %non-EU on the neighbourhood-level increases the odds of leaving for Swedish
residents by the factor 1.1 in 1991 in comparison to Swedish residents that do not experience the increase. A
unit increase in %non-EU in 1991 increases the odds of leaving for EU immigrant residents by the factor 1.2
in comparison to EU residents that do not experience the increase.
viii We ran individual-level regressions at each time point, when individuals had already moved to the new
neighbourhood (i.e. for movers between 1991 and 1992 in 1992).
ix A multinomial multilevel model is appropriate here because we want to model the composition of the in-
mover population. The nesting is individuals in neighbourhoods. The use of multilevel models, taking group
indicators as dependent variables to model segregation processes, has been applied by Goldstein and
Noden (2003) and Leckie and Goldstein (2015) before. Further information on the modelling exercise are
given in the appendix, section A4.
x Butts (2007) proposes a model for allocation of people that respects some of the systemic constraints;
Koskinen, Müller, and Grund (2017) propose to achieve this through extending the stochastic actor-oriented