Natives’ opinions on ethnic residential segregation and neighbourhood diversity in Helsinki, Oslo and Stockholm

Abstract

Nordic countries rank high on measures indicating tolerant views on immigrants. Yet, ethnic residential segregation is stated as being a major social problem in these countries. Neighbourhood flight and avoidance behaviour among the native born could be a sign of less tolerant views on minorities, but could of course be restricted to native-born residents in areas of high-ethnic concentration. So far, no research in these countries has explicitly focused on the majority population’s view on segregation, and we know little about how native-born residents in different neighbourhood contexts view ethnic segregation or how own residential experience shapes decisions on staying or leaving; this paper aims to help fill this research lacuna. In a survey targeting 9000 native-born residents in three Nordic capital cities—stratified into neighbourhood movers and stayers and into neighbourhoods having different proportions of non-Nordic-born residents—we answer three questions: do native-born respondents prefer a neighbourhood ethnic mix? Do they see ethnic segregation as a problem? Do they prefer lower, current or higher shares of ethnic minorities in their own neighbourhoods?

Keywords:

• Ethnic segregation
• majority population
• opinions
• Helsinki
• Oslo
• Stockholm

Introduction

Ordinary residents collectively shape neighbourhood population mix and the future of segregation by taking decisions regarding where to move to and from. This is true for all residents, majority and minority people alike, but the behaviour of the former normally has a much greater impact on residential patterns: they are more numerous, have on average better information and more economic resources and can thus make freer residential choices. If majority populations judge the quality and characteristics of residential areas by taking their ethnic composition into account, for instance, by exercising flight or avoidance behaviour in relation to areas of immigrant concentration, segregation might be difficult to combat by means of urban and housing policy. This is one reason for our choice to approach native-born residents with questions concerning their opinions on ethnic residential segregation. Is it a problem, and if so, for whom? We think it is reasonable to assume that people’s views on this issue are directly related to their own residential experience and to whether or not they prefer that neighbourhoods be ethnically mixed. Is it the case that natives who reside in immigrant-dense neighbourhoods are more pro-mix than those living elsewhere, or are they in fact dissatisfied with the actual level of minority residents in own neighbourhoods? This paper deals with these issues in the context of Nordic capital cities.

Neither of these questions has – as far as we can ascertain – been much researched in Europe and certainly not in northern Europe. Surveys carried out in some countries, such as by the Dutch Social and Cultural Planning Office (2004), repeated studies by Westin (e.g. 1984, 1987; Westin & Lange, 1993) and the Swedish Diversity Barometer (Mella & Palm, 2005) often include questions regarding tolerance for immigrant neighbours but do not analyse responses taking the respondents’ neighbourhood context into account. Until recently, existing research bringing in this local contextual dimension was heavily dominated by US experiences and focuses to a great extent on the black-white ethnic dichotomy (see, for instance, Farley et al., 1978, 1997). Later, attempts were made to broaden the perspective to include composition preferences in a multiethnic setting (see Zubrinsky Charles, 2000, for a study on Los Angeles), a situation at least to some extent more similar to the European one.

Besides surveying new aspects, we also intend to contribute to the research field of ethnic residential segregation, by focusing our attention not on ethnic minority residents but on the native-born majority. We share the position emphasized by Crowder et al. (2011, p. 41): ‘While scholarly attention continues to focus on immigrants’ residential attainment processes, we call for additional attention to native-born populations’ mobility reactions to these immigrant settlement patterns’. Some scholars choose to make such contributions by using register/census-based (Andersson, 2013; Kaufmann & Harris, 2015) or other secondary statistical sources (like the US panel data used by Crowder et al.), while the study presented here is based on a large-scale survey conducted in Helsinki, Oslo and Stockholm. Three thousand native-born neighbourhood movers and stayers in child-rearing ages (25–50) were sampled from the population registers and targeted in each city by a jointly developed questionnaire sent out by respective statistical authorities in the three countries.

The following section presents some earlier findings concerning the role that neighbourhood and segregation might have for people’s opinions and attitudes towards immigrants and neighbourhood diversity. The subsequent section outlines the empirical study with focus on sample design and data collection. Following that, we present results and deal with three main issues: whether or not respondents living in different diversity contexts prefer a neighbourhood ethnic mix, whether or not they see ethnic segregation as a problem and whether they favour lower, current or higher shares of ethnic minorities in their own neighbourhoods. Most data are presented using descriptive techniques (tables and diagrams), but we employ multivariate regression models for estimating which individual and neighbourhood characteristics contribute to explaining differences in stated preferences for the share of ethnic minority residents in one’s own neighbourhood. Section 5 summarizes the findings.

The role of neighbourhood in affecting attitudes, tolerance and trust

There is a long scholarly tradition, particularly based on intergroup contact reasoning, which theorizes the geography of tolerance. In American political science research on voting, the term racial threat refers to the association between white people’s propensity to vote for racial conservatives and the proportion of African-Americans living in their local area (Blalock, 1967; Key, 1949). This idea takes its point of departure in the fact that white racial attitudes and discriminatory behaviour are associated with the relative size of minority groups in a community. The idea that people’s social relations and behaviour are affected by racial/ethnic composition has much later taken centre stage in much of the ongoing social science research on social cohesion, trust and participation (Alesina & La Ferrara, 2000, 2002; Putnam, 2007), and it has been stated that ethnic diversity in communities and neighbourhoods undermines trust between people. In a series of papers (summarized in Uslaner, 2012), Uslaner argues that it is not diversity but social isolation produced by residential segregation that drives such results: ‘The problem is not diversity, but residential segregation. Living in segregated neighborhoods reinforces in-group trust at the expense of out-group (generalized) trust’ (p. 36).

In a recently published critical re-examination of Putnam’s study of diversity and trust, Abascal & Baldassari (2015) argue that his conclusion is a compositional artefact and that separate analyses need to be conducted according to ethnoracial groups. After having done so, they conclude: ‘Only for whites does living among outgroup members – not in diverse communities per se – negatively predict trust’ (p. 722). Other types of methodological criticism have also been launched in this Putnam-related debate, such as the Laurence & Bentley (2016) argument that cross-sectional research is inadequate if causal claims about the relationship are to be made. They call for a longitudinal test of the impact of diversity on social cohesion and attitudes towards neighbours, and they decompose residents along the stayer–mover dimension (see also Kaufmann & Harris, 2015). By splitting our respondents along the mover–stayer dimension (see next section), we are able to cast further light on this issue.

Reading the literature on attitudes towards ethnic minorities and some of the other partly related bodies of literature around issues of tolerance and trust, as well as literature on neighbourhood satisfaction, it is clear that the actual neighbourhood context of individuals matters, and so does ethnic residential segregation. In social psychology, Allport’s (1954) intergroup contact hypothesis has spurred hundreds of papers (Pettigrew & Tropp, 2008). It contends that intergroup contact typically diminishes intergroup prejudice. In a meta-analysis of more than 500 papers, Pettigrew & Tropp (2006) find that greater intergroup contact corresponds with lower levels of intergroup prejudice, and in their 2008 follow-up paper they show that it is not so much increasing knowledge about the outgroup that matters, but rather reducing anxiety and increasing empathy that mediate the negative relationship between contact and prejudice.

In an attempt to reconcile the contradiction found in previous research, that is, that the contextual contact literature finds that increasing the proportion of minority residents brings out negative majority attitudes toward minorities while the personal contact literature shows the reversed relationship, Stein et al. (2000) are able to identify a curvilinear relationship between the two dimensions: ‘Ironically, our findings suggest that the very conditions [increasing minority density] that give rise to white hostility toward minority group members set in motion a corrective for this hostility: inter-group contact’ (p. 299). For intergroup contact to occur, geographical proximity is necessary – be it in workplaces, neighbourhoods or elsewhere, where people may interact face to face.

Geographical proximity promotes intergroup contact, but it could also promote racial hostility. According to Quillian (1996), group threat theory proposes ‘that prejudice is in part a response to feelings that certain prerogatives believed to belong to the dominant racial group are under threat by members of the subordinate group. The central idea is that attitudes toward the other race are influenced by individual fears that their own race will be put at systematic disadvantage to the other race – it is thus group threat toward the racial group, not individual threat, that this theory emphasizes as a source of racial hostility’ (Quillian, 1996, p. 820).

Our research is not aiming to measure majority intergroup attitudes toward minorities per se but people’s view on ethnic segregation and diverse neighbourhoods, and the share of immigrants in one’s own neighbourhood might of course correlate with such attitudes. Furthermore, we do not engage in studying the role of direct intergroup contact but rather in studying the role of context in the way discussed by Stein et al. (2000) and Quillian (1996).

The neighbourhood satisfaction literature provides some further guidelines for how we approached our research questions and designed our survey. This research shows that perceptions of housing and neighbourhood rather than ‘objective’ neighbourhood characteristics are keys for understanding whether people are satisfied or not and whether they tend to stay put or move away (Lu, 1999; Permentier et al., 2011). Whether dissatisfaction with neighbourhood ethnic composition is an issue in this respect is a relevant question, but evidence on the effect of ethnicity on neighbourhood satisfaction is certainly mixed. Some studies show that neighbourhood satisfaction is strongly related to ethnicity (Campbell et al., 1976; Lu, 1999), while other studies find no effect (Harris, 2001; Parkes et al., 2002). Some studies (see Ellen, 2000) conclude that growing ethnic minority presence in a neighbourhood is associated with perceptions of structural decline, making it difficult to know whether ethnic composition in itself drives migration decisions.

Drawing upon the notion of residential stress, Feijten & Van Ham (2009) studied whether neighbourhood change – including increasing presence of non-Western immigrants – in the Netherlands affected the probability that residents wished to leave their neighbourhood. They found that ethnic change played a role, but the relationship turned statistically insignificant after they controlled for the residents’ subjective opinions about such change.

Existing research from the tolerance and social cohesion literature as well as from the neighbourhood satisfaction literature – albeit focusing on different issues and emanating from different social science disciplines – agrees that segregation and neighbourhood composition play a role in shaping people’s experiences and perceptions, which in turn might affect residential sorting across neighbourhoods. We summarize our reading and preunderstanding of segregation in the Nordic context to generate a set of expectations concerning results from our own study. These relate to the three research questions stated in the introduction.

(1)	Given the prominent position but negative connotations that ethnic residential segregation has in each Nordic country’s political and cultural debate (Andersson, R., et al., 2010), we expected that most of our native-born respondents in all three capital cities would regard ethnic segregation as a problem. We think it is natural to expect that people living in the most immigrant-dense neighbourhoods would regard the problem to be greater, not least because they would potentially face territorial stigmatization, that is, that their neighbourhood would have a bad reputation and be regarded as unattractive. Such a stigma would not only affect immigrants residing there, but native-born residents as well. Another reason for hypothesizing this involves selection effects due to sorting: natives staying in immigrant-dense neighbourhoods tend to have individual characteristics that are often found to co-vary with less positive views on immigrants, that is, they have on average less education and lower income (Quillian, 1996). People having recently left an immigrant-dense area have neighbourhood experiences similar to those of the stayers, but studies analysing selective migration in Swedish neighbourhoods show that out-movers have higher employment rates and higher incomes (Andersson & Bråmå, 2004; Andersson & Hedman, 2016). And while native avoidance behaviour gets support in housing mobility studies, native flight gets less support (Andersson 2013). We have no strong indication that the out-movers from immigrant-dense areas would have left simply because of a more critical view of diverse neighbourhoods, but our sample stratification design allows us to study this issue.
(2)	It is likely the case that a person’s own exposure to immigrants affects his or her view on ethnic residential segregation. However, without good network data on interpersonal contacts, hypothesizing precisely how is difficult. If the Swedes’ overall more positive attitudes towards immigration and immigrants are due to higher exposure to immigrants, we would expect Swedes to be more positive in relation to, for instance, neighbourhood ethnic mix. Analogous with this, we could expect respondents living in immigrant-dense neighbourhoods to be more favourable than others to neighbourhood ethnic mix. However, group threat theory and US studies focusing on the role of ethnic/racial context for attitudes toward minorities suggest the opposite outcome.
(3)	We expect one’s own neighbourhood’s ethnic composition to affect not only one’s view on whether segregation is a problem but also whether one wishes to see changes in the neighbourhood’s ethnic composition. Drawing upon research on intergroup contextual contact, we find it reasonable to hypothesize that natives residing in the most immigrant-dense areas are more inclined to state that they prefer a lower share than those residing elsewhere. Whether this is a linear relationship is also of great interest: is it that the ‘ethnic component’ (either directly, or indirectly if used as a proxy for structural problems; see Ellen, 2000; Laurence, 2016) only is relevant in areas having a very high share of immigrants, or is it rather that the proportion of respondents stating they would prefer fewer immigrants in their neighbourhood increases in a linear fashion from low- to high-density neighbourhoods? Is it the case that Helsinki – having low to modest immigrant density even in high density areas compared to Oslo and Stockholm – will come out very different in this regard? Based on the available literature, we had no clear preunderstanding of these relationships, but our tentative guess was that the reference point for most people is ‘normality’, and that normality is one’s own city.

Data collection and sample design

The surveys targeted native-born residents in child-rearing ages in the Helsinki, Oslo and Stockholm regions. They were carried out as postal surveys, with an optional possibility for Helsinki respondents to respond online (web-based questionnaire). We found no differences in the Helsinki material between those responding online and others. The questionnaires used in the three cities were more or less identical, but each research team could also add questions that seemed to be important in one context but maybe not in the other two. Questions were developed in English in the spring and summer of 2011, and they were later translated into the respective languages (Norwegian for Oslo, Swedish for Stockholm, and Finnish and Swedish for Helsinki, officially a bilingual city). The design of the questionnaires was conducted in dialogue with the respective statistical authorities, which were also commissioned to carry out the following tasks.

(1)	Identify the population and draw the sample from the complete population registry.
(2)	Send out the questionnaires and reminders (three reminders for Helsinki and Stockholm; two for Oslo). At least one reminder contained a new questionnaire.
(3)	Collect questionnaires and transform data into electronic form.
(4)	Add register data for each individual respondent.
(5)	Merge survey and register data into one single data file.
(6)	Carry out the non-response analysis and produce sample weights to be used in the analyses.
(7)	Produce the technical report and deliver the completed data file.

Questionnaires were sent out in September 2011 (Helsinki and Stockholm) and October 2011 (Oslo). The collection of questionnaires went on until around Christmas (January for Oslo), and the final data-sets were delivered in January to March 2012.

Four strata

The sample design is defined by (a) movers versus stayers, and (b) the share of non-Nordic immigrants in the neighbourhood (i.e. the proportion born outside of the Nordic countries). The samples were drawn from four subpopulations (in the complete population registries) residing in each capital region 1 January 2008 and still living there at the time the sample was drawn (summer 2011). In all three cities, 750 native-born residents aged 25–50 in 2008 in each of the following four strata received the questionnaire:

Stratum 1: ‘Stayers’ in the most immigrant-dense neighbourhoods. This subpopulation lived in any of the neighbourhoods belonging to the upper decile in terms of percentage of non-Nordic born residents (equal number of neighbourhoods in each decile). We labelled these neighbourhoods ‘Decile 10 neighbourhoods’, and Stratum 1 residents were labelled Decile 10 stayers. People in this category had lived in the same Decile 10 neighbourhood during 2008 and 2009 (two years) and were still living there at the time of the survey.

Stratum 2: ‘Movers’ out of a Decile 10 neighbourhood. A mover was defined as a resident who had lived in the same Decile 10 neighbourhood during 2008 and 2009 but had moved to another neighbourhood within the region in 2010 (and had not moved again thereafter). We labelled this subpopulation Decile 10 movers. A Decile 10 mover normally left this neighbourhood segment but could still be residing in high immigrant local density if moving into – for example – a Decile 9 neighbourhood.

Stratum 3: ‘Stayers’ in other types of neighbourhoods within the region (i.e. Decile 1 to 9 neighbourhoods). The basic definition of a stayer is the same as for Stratum 1. This category was labelled ‘Other stayers’.

Stratum 4: ‘Movers’ out of other neighbourhoods (i.e. Decile 1 to 9 neighbourhoods). We applied the same basic definition of a mover as for (2), and we labelled this category ‘Other movers’.

Table 1 shows some key data on the density of non-Nordic residents across deciles in the three cities. Although the overall immigrant density in the metropolitan region is highest in Stockholm, selection criteria differed with respect to whether second-generation immigrants were included (Oslo) or not (Helsinki and Stockholm) in the calculations. At the time of the study, Statistics Norway normally applied a wider definition of immigrants (including children of immigrants born in Norway) compared to Sweden and Finland but has recently changed into a more strict country of birth definition, The geographical definition also differed somewhat. In Helsinki, the targeted population lived in three municipalities (Helsinki, Vantaa and Espo); in Stockholm we applied Stockholm County (26 municipalities) as our geographical delimitation, while in Oslo we used Oslo and 12 surrounding municipalities in the county of Akershus. Altogether, this resulted in a higher proportion of non-Nordic residents in Oslo, with a neighbourhood average of 19.3 per cent, compared with 13.8 per cent in Stockholm and 6 per cent in Helsinki.

Table 1. Proportion of non-Nordic residents in neighbourhoods in Helsinki, Oslo and Stockholm.^a

Deciles of per cent non-Nordic in n’hoods	Helsinki	Oslo	Stockholm
10	10.9–20.9	39.4–64.2	24.3–66.7
9	8.3–10.8	27.0–39.0	17.5–23.9
8	7.0 –8.3	21.0–26.5	13.0–17.4
7	5.9–7.0	17.1–20.6	11.1–13.0
6	5.0–5.8	14.2–16.9	9.5–11.1
5	4.4–5.0	12.5–14.1	8.0–9.5
4	3.6–4.4	10.8–12.4	6.7–8.0
3	3.2–3.6	9.6–10.8	5.5–6.6
2	2.5–3.2	8.3–9.4	4.2–5.5
1	0.5–2.5	5.2–8.3	1.3–4.1
N’hood mean	6.0	19.3	13.8
N’hood median	5.0	14.1	9.5
Std. dev.	3.6	13.2	14.6
N	243	184	816

^a The Oslo values include second-generation immigrants, while Helsinki and Stockholm values are based on country of birth (first generation only). As a rule of thumb, the proportion of immigrants born in Africa and in Muslim countries increases in all three cities as the proportion non-Nordic residents grow.

Abbreviations: N’hood – neighbourhood; std. dev. – standard deviation.

In all cities, we furthermore excluded neighbourhoods having few residents (pop. under 500 in Oslo, 300 in Helsinki and 100 in Stockholm). The difference here has primarily to do with the size of the statistical areas, these being larger in Oslo (average size 5400 residents) and Helsinki (3900) compared to Stockholm (2100). When the questionnaires were sent out, it transpired that a few individuals in all three cities did not actually meet the basic criteria of being part of a subpopulation; they might have died, left the region or left the country. This ‘over-coverage’ reduces some of the samples somewhat; see Table 2.

Table 2. Subpopulations, sample size and response rates for the survey.

City	Stratum 1	Stratum 2	Stratum 3	Stratum 4	Total
	Decile 10 n’hood stayers	Decile 10 movers	Other stayers	Other movers
Oslo
Population	14 728	2931	2 03 199	33 954	2 54 812
Original sample	750	750	750	750	3000
Over-coverage	0	2	0	0	2
Net sample	750	748	750	750	2998
Responses	409	362	410	385	1566
Response rate	54.4	48.4	54.7	51.3	52.2
Helsinki
Population	29 102	2909	1 95 195	20 490	2 47 696
Original sample	750	750	750	750	3000
Over-coverage	1	3	0	1	5
Net sample	749	747	750	749	2.995
Responses	309	335	350	345	1339
Response rate	41.2	44.7	46.7	46.0	44.6
Stockholm
Population	24 826	346	3 51 158	37 794	4 17 238
Original sample	750	750	750	750	3000
Over-coverage	2	5	1	4	12
Net sample	748	745	749	746	2.988
Responses	298	321	396	356	1371
Response rate	39.8	43.1	52.9	47.7	45.9

Abbreviation: n’hood – neighbourhood.

As can be seen in Table 2, response rates were somewhat higher in Oslo than in Stockholm and Helsinki but we found response rates across the three cities and the four strata in each city to be acceptable, or even good by contemporary survey standards. We also have a good picture of the non-responses. To summarize: in all surveys, targeted participants in some categories are less inclined to respond. Broadly speaking, males, young people, people of low income and those having a low level of education have lower response rates than their demographic and socioeconomic opposites. This was the case in this survey, and such differences also explain most of the differences in response rates across the four strata in each city. The order for Helsinki and Stockholm was that ‘Other stayers’ had the highest response rate, followed by ‘Other movers’, ‘Decile 10 movers’ and finally, ‘Decile 10 stayers’. Oslo showed the highest response rates for the two stayer categories.

We mentioned above that the statistical authorities carrying out the data collection also added register data to the respondent files. We therefore did not have to ask about the respondents’ age, gender, family type, income and education (education was not included for Stockholm). In the Stockholm and Oslo cases, we also had the specific neighbourhood codes for 2008 to 2011, which made it possible to add other neighbourhood characteristics in the multivariate models for these two cities. When responding to the questionnaire, the participants agreed to register information being used in the analyses.

For most parts of the following data analyses we kept the three data-sets apart, carrying out identical but separate analyses for each city. However, to make full use of the comparative potential, we merged the three data-sets into one when carrying out one final multivariate analysis. The unweighted combined material comprises 4276 respondents. They represent more than 900 000 Helsinki, Oslo and Stockholm residents in the selected age group. We used the city-specific post-stratified weights provided by the statistical authorities (taking account of differences in response rates), and when applying these in descriptive analyses reported in tables and figures, there is no need to present confidence intervals (very narrow confidence spans).

Key features of neighbourhoods in the three cities

It is far beyond the scope of this paper to give a comprehensive account of the historical and geographical dimensions of immigration and housing development in the three cities under study. Other parts of our comparative Nordic project provide such overviews (see, for example, country overviews in Andersson, R., et al., 2010; Skifter Andersen et al., 2016; Wessel et al., 2016). This section provides some basic data concerning the demographic and socio-economic structure of neighbourhoods having different proportions of non-Nordic immigrants.

As can be seen in Table 3, Helsinki has the lowest proportion of non-Nordic residents but also experienced the highest relative increase in the 2000s (comprising in particular immigrants from Russia, Estonia and Somalia), a fact that has triggered substantial interest among politicians and planners regarding the development of ethnic residential segregation. Such an interest has a longer history in Oslo (where the largest non-Nordic countries of origin are Poland, Lithuania, Pakistan, Somalia and Iraq), and especially in Stockholm (the largest groups being from Iraq, Poland, Iran, Turkey and Chile). Both these cities have also experienced substantial growth in the non-Nordic population and have launched area-based urban programmes with the aim to combat segregation and make immigrant-dense areas more attractive (Andersson 2006; Andersson, R., et al., 2010; Ruud et al., 2011).

Table 3. Overview of neighbourhood group characteristics in Oslo, Stockholm and Helsinki (groups according to deciles of neighbourhoods based on the proportion of non-Nordic residents).

	Non-Nordic decile 1 to 3	Non-Nordic decile 4 to 6	Non-Nordic decile 7 to 9	Non-Nordic decile 10	Total		Non-Nordic decile 1 to 9
	Mean	Mean	Mean	Mean	Mean	Std. dev.	Mean	Std. dev.
Oslo
Average population	4400	5004	5721	5970	5129	20 199	5038	20 453
%Non-Nordic pop in n’hood (only 1st generation)	9.0	12.7	22.1	45.9	17.6	11.3	14.6	6.4
%Change in non-Nordic pop 2000 to 2008	72.0	61.2	60.8	65.1	64.7	42.0	64.7	42.8
%Non-Nordic children age 0 to 17	10.4	14.8	31.1	63.0	23.0	16.8	18.7	10.6
%High education	36.9	45.4	39.5	21.9	38.8	13.1	40.6	12.3
%Low education	22.6	18.4	22.4	35.4	22.5	7.7	21.1	6.5
%Public rental housing	1.9	1.9	5.6	6.4	3.5	4.1	3.1	3.7
%Total rental	13.3	15.8	26.9	20.8	18.8	10.9	18.6	10.7
%Homeownership	77.4	68.3	40.0	30.8	58.9	26.1	62	24.8
%Cooperative housing	9.2	15.8	32.7	47.9	22.0	23.3	19.2	21.3
Stockholm
Average population	1243	2639	2810	4248	2426	27 718	2225	2 60 930
%Non-Nordic pop in n’hood (only 1st generation)	4.6	8.8	15.8	39.0	12.6	10.6	9.7	5.22
%Change in non-Nordic pop 2000 to 2008	36.9	36.2	45.6	40.8	39.6	62.0	39.5	64.57
%Non-Nordic children age 0 to 17	3.6	8.3	20.3	62.3	15.9	18.6	10.7	9.45
%High education	22.0	30.6	25.5	15.4	24.9	12.3	26.0	12.30
%Low education	40.8	30.0	35.8	50.9	37.1	13.6	35.6	13.04
%Public rental housing	1.0	6.5	16.8	43.0	11.5	21.7	8.0	16.88
%Total rental	3.0	14.6	34.2	61.7	21.5	28.7	17.1	24.28
%Homeownership	90.9	60.2	28.6	12.9	55.6	41.7	60.3	40.26
%Cooperative housing	6.1	25.2	37.2	25.4	22.9	28.2	22.6	28.02
Helsinki
Average population	2346	4153	4842	6087	4013	32 155	3775	31 730
%Non-Nordic pop in n’hood (only 1st generation)	2.8	4.7	7.8	14.4	6.0	3.6	5.1	2.20
%Change in non-Nordic pop 1999 to 2010	125.6	118.4	118.6	106.9	118.9	101.5	120.6	107.27
%Non-Nordic children age 0 to 17	2.1	4.6	10.7	23.1	7.5	7.2	5.8	4.72
%High education	48.2	45.8	37.3	25.8	41.8	11.9	43.7	10.92
%Low education	17.7	17.9	22.7	32.7	20.9	6.8	19.5	5.47
%Public rental housing	6.8	14.0	28.5	41.9	19.0	17.4	16.4	15.95
%Total rental	17.0	30.3	45.6	53.2	33.2	18.6	30.9	18.08
%Homeownership	72.3	58.6	42.6	38.6	55.9	19.6	57.9	19.48
%Cooperative housing	1.7	2.4	4.1	4.5	2.9	5.3	2.7	5.09

^a The Finnish data refer either to the year 2008 (end of year 2007 situation) or residency 31 December 2007 (change in non-Nordic population has been calculated from 31 December 1999 to 31 December 2010).

Notes: All percentages are calculated with total population in neighbourhood as a basis, or people aged 20 to 64 (education, work), and people over 15 (income).

Student campuses of Viikin tiedepuisto and Otaniemi, and all neighbourhoods with less than 300 residents (31 December 2007) have been excluded from the table.

Stockholm has the biggest difference between low and high immigrant-dense areas: while the Deciles 1–3 versus Decile 10 mean proportion is around 1 to 5 in Oslo and Helsinki, it is more than 1 to 8 in Stockholm. It should also be noted that the composition of non-Nordic residents in neighbourhoods changes in all three cities as the overall proportion of non-Nordic residents increases. A higher proportion of non-Nordic population typically means an even higher concentration of people originating in Muslim and African countries. This reflects an ethnic hierarchy in majority attitudes, labour market status and housing conditions, where those population segments run a much higher risk of being stigmatized and discriminated against. It could also be added that the variety of ethnic groups is high in Decile 10 neighbourhoods in all three cities (‘multi-cultural’) but due to fewer migrant groups represented in Helsinki, variety is less in this city. In all three cities, the most immigrant-dense areas have a very high proportion of children aged 0–17 with non-Nordic background, a fact that certainly indicates that school-related issues are an integrated part of the ethnic segregation debate in the cities (Andersson, E., et al., 2010; Trumberg, 2011). The proportion of highly educated residents (15 years or more of schooling) is generally higher in Oslo and Helsinki, but in each city the proportion is much lower in the Decile 10 neighbourhoods. It is normally the case that especially natives living in immigrant-dense areas have on average a low level of education. Finally, there are big differences between the cities when it comes to housing tenure structures. While Oslo, including the Decile 10 neighbourhoods, is heavily dominated by market-priced housing (ownership and cooperative housing), rental housing and especially social/public housing, is a key characteristic of immigrant-dense neighbourhoods in Stockholm and Helsinki. It is reasonable to expect these housing tenure differences to affect how residents view ethnic mix in neighbourhoods. Other studies have confirmed that owners are more sensitive to changes in ethnic composition and that they, earlier than renters, tend to exercise flight and avoidance behaviour in relation to increasing proportions of ethnic minorities (Bråmå, 2006; Ellen, 1997; Goering, 1978).

Results

Do native residents prefer neighbourhood ethnic mix?

In Stockholm, two-thirds of all respondents stated that they completely agree or agree to some extent with this proposition. The corresponding proportion for Helsinki is 43 per cent and for Oslo 53 per cent. The differences between the four basic strata within the cities are small, so at least when not controlling for other individual attributes, the preferences for ethnically mixed neighbourhoods seem not to depend much on people’s own residential context. This is apparent also if we apply more precise ethnic information to the respondents’ current neighbourhoods (i.e. post-migration for the two mover strata). Figure 1 shows the percentage stating a preference for ethnic mix by city and current (2011) neighbourhood decile of the proportion of non-Nordic residents. Firstly, the chart could be read as an illustration of the fact that Swedes (Stockholmers) in general report more positive attitudes towards immigrants and ethnic minorities. The proportion agreeing that ethnic mix is positive was higher in Stockholm in each of the 10 deciles, but the discrepancy is most pronounced in the bottom decile segments, that is, among those that reside in high-density ‘native areas’. Secondly, in Helsinki, less than half of the respondents seem to be pro-ethnic mix.

Figure 1. Proportion (in per cent) of residents stating a preference for ethnic mix in their neighbourhood (by current non-Nordic immigrant density in neighbourhood). Neighbourhood ethnic diversity increases from left to right.

Thirdly, in Oslo we discern a tendency for the ethnic mix idea to get somewhat more support in higher immigrant density than in low immigrant density neighbourhoods. Expressed in another way but without arguing direction of causality: Oslo residents having more experience of immigrants in their own neighbourhoods are more inclined to voice pro-ethnic-mix attitudes.

Even though the preference for neighbourhood ethnic mix seems to be widespread in Stockholm and is relatively common also in the other cities, many respondents also state that they prefer not to live close to certain ethnic groups. Here also, more – in fact a majority – of the Helsinki (61 per cent) and Oslo (53 per cent) respondents state this preference, while in Stockholm the proportion is below 40 per cent. In Table 4 we have combined this information by cross-tabulating statements concerning ethnic mix in neighbourhoods with the expressed reluctance to live close to certain minority groups. The relation between the cities is that Stockholm respondents are more likely than Helsinki respondents to favour a mix and to show tolerance of all groups. Oslo has an in-between position in this regard.

Table 4. Proportion of respondents per city that state different combinations of attitudes towards ethnic mix and willingness to live close to certain ethnic groups.

		Helsinki			Oslo			Stockholm
		Agree that certain groups should be avoided			Agree that certain groups should be avoided			Agree that certain groups should be avoided
		Yes	No	Total	Yes	No	Total	Yes	No	Total
Prefer ethnic mix	Yes	16.4	26.7	43.1	14.4	32.3	46.7	17.2	49.5	66.8
	No	44.2	12.7	56.9	37.2	16.1	53.3	19.9	13.3	33.2
	Total	60.6	39.4	100	51.6	48.4	100	37.1	62.8	100

Is ethnic residential segregation a problem or not and for whom?

Is ethnic segregation a problem? We asked our respondents their opinions on this issue. Table 5 shows clearly that an overwhelming majority of native Oslo and Stockholm respondents found ethnic residential segregation to be a problem, and a problem for natives as well as immigrants. The Helsinki outcome is slightly different, but very few stated that it is no problem or that they do not know whether it is a problem. For Oslo and Stockholm residents, the ‘two-sided problem’ opinion was slightly more common among those having left the most immigrant-dense areas (Stratum 2) than it was among the stayers in immigrant-dense areas. In all three cities, it was the case that stayers in the immigrant-dense neighbourhoods (‘Decile 10 stayers’) more often stated that ethnic segregation is a problem for natives. We cannot be sure this is not a result of selection of residents, but this tentatively confirms our expectation that people’s own local experience affects their views on segregation, including informing the way they perceive for whom it is a problem.

Table 5. Is ethnic residential segregation a problem? Answers by city and stratum.

Is ethnic residential segregation a problem?	DEC10 stayers	DEC10 movers	DEC1–9 stayers	DEC1–9 movers	Total
Helsinki
No	4.8	4.5	4.0	4.9	4.2
Yes, a problem for immigrants	1.1	2.6	2.8	2.1	2.6
Yes, a problem for Finns, but not for immigrants	12.4	11.0	6.9	4.0	7.4
Yes, a problem for both groups	36.1	36.8	36.1	38.9	36.4
Sometimes a problem, sometimes not	40.2	39.6	43.2	41.8	42.7
Do not know	5.5	5.5	6.9	8.2	6.9
Total	100.0	100.0	100.0	100.0	100.0
Oslo
No	3.6	2.8	2.8	4.8	3.0
Yes, a problem for immigrants	5.4	4.0	5.6	7.2	5.7
Yes, a problem for Norwegians, but not for immigrants	7.8	4.0	4.7	3.4	4.8
Yes, a problem for both groups	59.7	64.1	58.0	56.1	58.0
Sometimes a problem, sometimes not	18.2	21.3	22.4	22.4	22.2
Do not know	5.3	3.7	6.5	6.2	6.4
Total	100.0	100.0	100.0	100.0	100.0
Stockholm
No	4.5	3.0	4.0	3.7	4.0
Yes, a problem for immigrants	6.0	5.9	4.5	4.3	4.5
Yes, a problem for Swedes but not for immigrants	5.5	2.0	1.5	1.5	1.7
Yes, a problem for both groups	48.8	55.9	59.7	61.6	59.2
Sometimes a problem, sometimes not	28.3	28.4	24.4	26.2	24.8
Do not know	6.9	4.8	5.9	2.7	5.7
Total	100.0	100.0	100.0	100.0	100.0

Abbreviation: DEC – decile.

Most of the segregation discussion in the media tends to focus attention on specific city districts and neighbourhoods, that is, localizing the segregation issue to particular areas and disregarding the fact that spatial concentration of a particular group is possible only if other areas have less representation of the same group. The focus on ‘immigrant-dense neighbourhoods’ is also produced by actual policies. It is common for central governments and municipalities to launch area-based urban policies, a type of intervention that most often targets precisely these types of neighbourhoods (see Andersson, 2006; Andersson & Musterd, 2005). It is therefore not surprising that around one out of four respondents in Stockholm, one out of five in Oslo and two out of five in Helsinki mentioned this aspect when asked about segregation as a problem (‘segregation a problem in some areas but not in others’). One possible explanation for the high share in Helsinki could be that ethnic minorities are more concentrated-albeit at lower levels of concentration – in Helsinki (fewer areas connected to the notion of immigrant-dense areas) and that more people therefore tend to localize the phenomenon.

Are people satisfied with own neighbourhood’s ethnic composition?

Expressing a preference for a lower share of immigrants is the plurality category only for those residing in a Decile 10 neighbourhood. Overall, Oslo residents voiced more concern over the immigrant presence in their own neighbourhoods, and this was especially true of those living in Decile 10 neighbourhoods (see Figure 2). Two out of three (66 per cent) of these residents stated they would prefer lower immigrant presence in their neighbourhood. The corresponding value for Stockholm Decile 10 residents was 47 per cent and for Helsinki 41 per cent. Furthermore, in Stockholm a relatively big proportion of non-Decile 10 respondents declared that they would prefer a higher share, and although most said they preferred the current level, many more stated higher compared to lower. In Oslo, and even more clearly in Helsinki, those wanting a higher proportion of immigrants were less common among our respondents.

Figure 2. Respondents according to subsample who prefer higher, lower or current share of immigrants in neighbourhood.

Note: DEC = decile.

It is important to note that the stated preference of Decile 10 movers should be seen in relation to the post-migration situation (i.e. when responding to this question most no longer lived in a Decile 10 neighbourhood). We have therefore calculated also the immigrant density of the respondents’ current (2011) neighbourhood, and in analogy with the definition of our samples, we have re-sorted them into their current non-Nordic deciles. Figure 3(a)–(c) displays the proportion of people by current decile position who preferred lower, current and higher presence of immigrants in their neighbourhood.

Figure 3. Respondents according to current neighbourhood immigrant density who would prefer higher, lower or current share of immigrants in neighbourhood.

A couple of clear messages emerge from Figure 3. Most respondents preferred the current share of immigrants, irrespective of their own current neighbourhood context. There was, however, one exception, and that was those residing in a Decile 10 neighbourhood. Here, the plurality category comprises those preferring a lower share. Furthermore, the proportion of respondents wanting to see a higher share is inversely related to immigrant density: The higher the proportion of immigrants, the lower the proportion of respondents wanting a higher share. Finally, and not least notable, there seems to exist a threshold over which the proportion of respondents stating a preference for a lower share of immigrants starts to increase rapidly. That threshold is around the 7th decile in Stockholm, and the 8th decile in Helsinki and Oslo. In all three cities, this breakpoint is around the city average level of percentage of non-Nordic immigrants (see Table 1), confirming our tentative expectation that one’s frame of reference is the normality of one’s own city.

It seems obvious that the view on immigrant density is context dependent, confirming not only the general relevance of the intergroup contextual contact theory but also the need to acknowledge intercity variations in proportion of immigrants. Not only does the actual level of immigrant presence affect people’s preferences but it is also the case that the profiles across the three cities seem to vary much less than one might expect. While, for example, Decile 8 means about 7 to 8 per cent non-Nordic immigrants in a Helsinki neighbourhood (see Table 1), it means twice as many in Stockholm and (including second generation) three times as many in Oslo. Still, about 20 per cent of the Decile 8 respondents in all three cities stated that they want to see a lower share of immigrants in their neighbourhood. We believe this to be an important result. People seem to view low and high shares in relation to the city context. A high level by Helsinki standards therefore triggers similar sentiments and opinions regarding immigrant density as it does in Oslo and Stockholm. Increasing the share of immigrants in Helsinki or lowering the share in Oslo and Stockholm will probably not affect this preference structure.

Multivariate modelling

Of course, a person’s neighbourhood context is but one of many factors that may affect his or her view regarding lower/higher share of immigrants in one’s own neighbourhood. The literature referred to earlier in this paper suggests that social background influences people’s perceptions of immigrants; residents with high social and economic capital can be expected to be more tolerant, irrespective of their neighbourhood location. To get a clearer picture of such more complex associations, we have modelled one of the preferred alterations in neighbourhood composition stated by the respondents: those wanting a lower share of immigrants in their neighbourhood.

We employed a set of control variables intended to capture variations in demographic attributes (gender, age and household composition) and socio-economic attributes (income and employment), as well as the individuals’ current neighbourhood context and tenure. Our primary contextual variable of interest was the proportion of non-Nordic residents in one’s own neighbourhood, but for Oslo and Stockholm we included three other (but partly different) contextual control variables: proportion employed/unemployed, proportion highly educated and income (proportion low or mean values). We also add some self-reported data (on the respondent’s view on safety, his/her social interactions with neighbours and participation in neighbourhood work/associations). US studies have repeatedly reported on and confirmed the close relationship between real or perceived feelings of safety and a neighbourhood’s racial and ethnic composition (Emerson et al., 2001; Quillian & Pager, 2001; Sampson & Raudenbusch, 2004). Research on participation in relation to European large housing estates – which in the Nordic context often means high levels of non-native-born residents – finds that differences in local activity levels have national, neighbourhood and individual determinants and that this could affect people’s views on issues related to neighbourhood satisfaction (Dekker & Van Kempen, 2008; Dekker et al., 2011).

Table 6 displays the mean and standard deviation values per city for all variables included in the regression model. The model was run both in SPSS, using the Complex Sample logistic regression option with city-scaled post-stratified weights as the plan scheme, and in Stata, using logistic regression with robust standard errors clustered on neighbourhoods. Results were very similar, and we report the Stata output.

Table 6. Descriptives of variables used in the logistic regression model.

	Descriptives, post-stratified weights
	Stockholm			Oslo			Helsinki			Total
Independent variable	Mean	N	Std. dev.	Mean	N	Std. dev.	Mean	N	Std. dev.	Mean	N	Std. dev.
Male	.51	1371	.50	.49	1566	.50	.48	1339	.50	.49	4276	.50
Under 35	.18	1371	.38	.15	1566	.36	.18	1339	.39	.17	4276	.37
Family with child(ren)	.69	1371	.46	.53	1566	.50	.46	1339	.50	.60	4276	.50
Disposable household income quintile 1 (lowest 20%)	.20	1371	.40	.08	1566	.27	.10	1339	.30	.13	4276	.33
Disposable household income quintile 2	.21	1371	.41	.21	1566	.41	.10	1339	.30	.18	4276	.38
Disposable household income quintile 3	.19	1371	.40	.18	1566	.38	.18	1339	.38	.18	4276	.39
Disposable household income quintile 4	.20	1371	.40	.26	1566	.44	.26	1339	.44	.24	4276	.43
Disposable household income quintile 5 (highest 20%)	.20	1371	.40	.27	1566	.44	.36	1339	.48	.27	4276	.45
Works full time	.80	1371	.40	.81	1566	.39	.82	1339	.39	.81	4276	.39
Lives in rental housing	.22	1371	.41	.05	1566	.22	.24	1339	.43	.17	4276	.37
Discontent with safety in current neighbourhood	.06	1371	.23	.06	1566	.24	.09	1339	.28	.07	4276	.25
Speaks to neighbours at least monthly	.89	1371	.31	.88	1566	.32	.88	1339	.33	.88	4276	.32
Visits at least monthly	.60	1371	.49	.59	1566	.49	.40	1339	.49	.54	4276	.50
Assists neigbours at least monthly	.36	1371	.48	.40	1566	.49	.36	1339	.48	.37	4276	.48
Participates in n’bourhood work	.12	1371	.30	.15	1566	.36	.13	1339	.33	.14	4276	.34
Mover	.49	1371	.50	.48	1566	.50	.51	1339	.50	.49	4276	.50
Lives in Non-Nordic Decile 1–2 in 2011	.11	1371	.31	.17	1566	.37	.11	1339	.32	.13	4276	.34
Lives in Non-Nordic Decile 3–7 in 2011	.64	1371	.48	.55	1566	.50	.52	1339	.50	.57	4276	.50
Lives in Non-Nordic Decile 8–9 in 2011	.19	1371	.39	.22	1566	.42	.24	1339	.43	.22	4276	.41
Lives in Non-Nordic Decile 10 in 2014	.06	1371	.24	.06	1566	.25	.13	1339	.33	.08	4276	.28
Percentage (aged 20 to 64) employed in n’hood in 2010	.77	1371	.09
Percentage unemployed in n’hood 2010				.05	1566	.01
Percentage (aged 20 to 64) highly educated in n’hood in 2010	.28	1371	.12	.35	1566	.13
Percentage (aged 20 to 64) with low income in n’hood 2010	.29	1371	.10
Average neighbourhood income in 2009 (1000 NOK)				281.0	1566	44.9
Dependent variable: prefer lower immigrant density in neighbourhood	.10	1371	.30	.16	1566	.37	.19	1339	.39	.15	4276	.36

The descriptives presented in Table 6 show that variations across the cities were relatively small for most variables, but some deviations from this pattern should be noted. Families with children made up a bigger share of the Stockholm population, people with high income composed a bigger share in Oslo and Helsinki compared to Stockholm and renters were few in Oslo. With respect to the self-reported statements, participating in neighbourhood work, speaking with neighbours and reporting neighbourhood safety concerns were of similar magnitude across the cities, while visiting neighbours on a regular basis seemed to be more common in Stockholm and Oslo than it is in Helsinki. When it comes to the dependent variable (expressing preference for lower share of immigrants in one’s own neighbourhood), we have already reported above that the Stockholm residents less frequently stated this preference.

For Oslo and Stockholm, where we had neighbourhood codes, we ran the models with robust standard errors clustered on neighbourhoods. This was not possible for Helsinki and therefore not included in the joint model run, but it should be said that we did test run Oslo and Stockholm without the clustering on neighbourhood option and with almost identical outputs (likely because of few respondents per individual neighbourhood, so clustering means little).

Table 7 displays odds ratios for model runs per city, and in the final column for the joint run. The joint run contained a city control, and this confirms that it was almost twice as likely that an Oslo or a Helsinki respondent stated a preference for a lower share of immigrants compared to a Stockholm respondent.

Table 7. Logistic regression outputs. Dependent variable: prefer lower share of immigrants in own neighbourhood.

		Dependent variable: prefer lower share of immigrants in neighbourhood
		Oslo			Stockholm			Helsinki			All three cities
Variable type		Robust standard errors clustered on neighbourhoods			Robust standard errors clustered on neighbourhoods			Robust standard errors			Robust standard errors
	Independent variables	B	P > |z|	Odds Ratio	B	P > |z|	Odds Ratio	B	P > |z|	Odds Ratio	B	P > |z|	Odds Ratio
Sex	Males	.416	.089	1.516*	.793	.013	2.209**	.309	.194	1.362	.493	.002	1.638***
	Females (Ref.)
Age	Age under 35	−.999	.015	0.368**	.937	.051	2.553*	−.228	.529	.796	−.145	.557	.865
	Age over 34 (Ref.)
Family	Family with children aged 0–17	−.160	.578	.852	.366	.290	1.442	−.050	.846	.951	.120	.462	1.128
	Family without children aged 0–17 (Ref.)
Income	Disposable income, lowest quintile (Ref.)
	Disposable income, quintile 2	.244	.576	.576	.282	.545	1.326	.415	.333	1.514	.203	.427	1.226
	Disposable income, quintile 3	.173	.735	.735	.647	.155	1.909	.820	.048	2.284**	.443	.080	1.557*
	Disposable income, quintile 4	−.074	.877	.877	.717	.180	2.048	.561	.180	1.752	.247	.330	1.280
	Disposable income, highest quintile	−.226	.667	.667	.616	.318	1.851	.271	.548	1.311	−.040	.888	.961
Employment	Work full time	−.709	.048	.492**	.557	.180	1.746	.527	.139	1.694	.118	.580	1.125
	Does not work full time (Ref.)
Tenure	Rental	−.568	.295	.566	.576	.138	1.778	.191	.522	1.211	.164	.415	1.179
	Not rental (Ref.)
Self-reported statements	Speak with n’bours at least monthly	−.606	.132	.545	−.387	.418	.679	.797	.038	2.218**	−.166	.474	.847
	Speak with n’bours more seldom (Ref.)
	Visit neighbours at least monthly	−.120	.689	.887	.408	.260	1.504	−.098	.693	.906	.019	.912	1.019
	Visit n’bours more seldom (Ref.)
	Assists or is assisted by neighbour at least monthly	−.562	.121	.570	−.688	.065	0.503*	−.283	.291	.754	−.468	.007	0.626***
	Does not assist neighbours montly (Ref.)
	Participate monthly in neighbourhood associations	−.038	.931	.963	−.066	.920	.936	.479	.158	1.614	.107	.647	1.113
	Does not participate monthly (Ref.)
	Discontent with safety	1.957	.000	7.081***	1.729	.000	5.632***	1.531	.000	4.620***	1.617	.000	5.040***
	Does not state discontent with safety
Stayer/mover	Stayer (Ref.)
	Mover	.303	.196	1.355	.022	.945	1.022	−.213	.337	.808	−.213	.158	.808
Neighbourhood ethnic context	Non-Nordic Decile 1–2 (Ref.)
	Non-Nordic decile 3 to 7	1.423	.148	4.151	1.813	.122	6.127	.023	.960	1.024	.577	.11	1.781
	Non-Nordic decile 8–9	2.487	.016	12.023**	2.272	.051	9.698*	.975	.047	2.650**	1.748	.000	5.746***
	Non-Nordic decile 10	3.744	.001	42.276***	3.000	.016	20.087***	1.434	.001	4.196***	2.549	.000	12.790***
Extra neighbour-hood controls for Oslo and Stockholm	% Employed in n’hood aged 20 to 64 (Stockholm)				−.125	.495	.883
	% Unemployed in n’hood (Oslo)	−.040	.820	.960
	% Highly Educated (15 years or more)	−.038	.034	0.963**	−.096	.000	0.908***
	% Low income (lowest 3 work income deciles) Stockholm				−.111	.498	.895
	Mean income in n’hood (1000 NOK)	.002	.808	1.000
City	Stockholm (Ref.)
	Helsinki										.625	.001	1.868***
	Oslo										.668	.000	1.951***
Constant		−2.493	.002	.083***	−5.381	0.000	.005***	−3.879	.000	.021***	−3.800	.000	.022***
Pseudo R^2		.263			.335			.137			.191
No. of neighbourhoods		178			473

Note: P > |z|.

^* p <0.10

^** p < 0.05

^*** p < 0.01.

Furthermore, Table 7 shows that some demographic and socio-economic characteristics included in the models are not in a statistically significant way related to stating a preference for a lower share of immigrants. While gender and age are significant in Oslo and Stockholm but not in Helsinki, and not in the joint model, having children or not, renting or not and having a high or low income seem not to matter for expressing this preference. Moreover, we found no difference between movers and stayers in any of the three cities. We do not know why this is the case but it confirms earlier studies that ‘white flight’ (the ‘ethnic component’) is not a strong driver for out-migration from immigrant-dense neighbourhoods but rather socioeconomic factors (Andersson 2013). Working full time is negatively related to stating a preference for a lower share of immigrants in Oslo but not in Stockholm or Helsinki or in the joint city run. The safety issue is strongly related to a preference for a lower share of immigrants: voicing concern regarding safety (reporting feeling unsafe) in one’s own neighbourhood strongly predicts a preference for a lower share of immigrants (five times more likely compared to not feeling unsafe).

The odds quotas for neighbourhood ethnic context in Table 7 are all related to the respective cities’ respondents living in low immigrant-dense neighbourhoods (Deciles 1 and 2). We found no statistically significant difference between people residing in low (Deciles 1 and 2) and other low to medium immigrant concentration (Deciles 3 to 7) neighbourhoods in any of the three cities. However, the earlier finding that Deciles 8 and 9, and especially Decile 10, residents were far more inclined to state a preference for a lower share of immigrants is very robust even after controlling for other individual characteristics, and in the case of Oslo and Stockholm also after controlling for some other key neighbourhood characteristics.

We can summarize these findings in a few words. Firstly, our literature-derived expectation that the neighbourhood location of individuals matters for their stated preferences was confirmed. Secondly, irrespective of neighbourhood ethnic context, we found that Stockholm respondents reported more positive attitudes vis-à-vis having immigrant neighbours. Finally, we found that people seem to view low and high shares of immigrants in relation to the city context. Despite the fact that high immigrant density represents a relatively low concentration in Helsinki compared to Stockholm and Oslo, a high level in Helsinki triggers sentiments and opinions regarding immigrant density similar to those found in the other cities. This relational dimension is important to acknowledge for achieving a better understanding of popular imaginations and preferences.

Conclusions

International comparative research has repeatedly found that the Nordic countries score high on welfare indicators. Class differences are relatively modest, the living standard is high and the universal approach to public service delivery provides people with a high degree of safety when falling sick or being unemployed or retired. The countries come out as ‘high trust societies’ measured as both in- and outgroup trust (Delhey & Welzel, 2012). Studies of tolerance for ethnic minorities also rank the Nordic countries high, but with more variation. Normally, surveys indicate that Swedes are the most tolerant and also most pro-immigration.

We do not dispute these findings, but following arguments provided by proponents of intergroup contextual contact theory (Quillian 1996; Stein et al., 2000), we have emphasized the importance of studying interethnic relationships and tolerance issues not only in a macrocontextual setting (comparing countries) but also by taking geographical microcontexts into account. As pointed out by Alesina & La Ferrara (2000), Putnam (2007), Feijten & Van Ham (2009), Uslaner (2012) and others, issues concerning social cohesion, trust and tolerance need to be understood in a community-oriented way, focusing on the reality that people face in their daily lives. In particular, Uslaner stresses the importance of ethnic residential segregation for understanding how in- and outgroup trust develops.

Besides taking these insights on board and following arguments put forward by Crowder et al. (2011) and by many scholars studying racism and discrimination (see, for instance, Andersson & Molina, 2003), we emphasize the role of the native majority population for understanding segregation processes and intergroup relations. Our study is relatively unique in a European context because it (a) surveys natives’ experiences of and perspectives on ethnic segregation, (b) provides a basic sample stratification that allows us to differentiate between neighbourhood stayers and movers and between those residing in different native/immigrant contexts and (c) surveys three Nordic capital regions, using the same questionnaire.

In the introductory part, we raised three research questions: Is ethnic residential segregation a problem, and if so, for whom? Are natives who reside in immigrant-dense neighbourhoods more pro-mix compared to those living elsewhere, or are they in fact dissatisfied with the actual level of minority residents in their neighbourhood? Do they prefer lower levels? This paper has tried to answer these questions in the Nordic capital city contexts of Helsinki, Oslo and Stockholm.

With reference to the negative connotations that ethnic residential segregation has in each Nordic country’s political and cultural debate, we expected most of our respondents in all three capital cities to view ethnic segregation as a problem. This expectation is confirmed by our study. Around 90 per cent of the respondents agreed that it is a problem, and most tended to view it as a problem for immigrants as well as natives. Respondents in Helsinki had a stronger tendency to locate the segregation problem in specific parts of the city than did the respondents in Oslo and Stockholm. We think that this has to do with not only the short immigration history but also the fact that – given the much lower overall presence of immigrants in Helsinki – many fewer neighbourhoods are affected by immigrant concentration, so it might be easier to relate ethnic segregation to a more narrow set of neighbourhoods or city districts.

We furthermore expected that people living in the most immigrant-dense neighbourhoods would regard the problem to be bigger, not least because they would potentially face territorial stigmatization, that is, that their neighbourhood would have a bad reputation and be regarded as unattractive. We did not find clear support for this hypothesis, and the reason might be that ethnic residential segregation is an established political and media issue. As such, people’s opinions regarding the phenomenon are less dependent on their own experience but rather are informed by these more general public discourses.

With respect to people’s views on ethnic mix in neighbourhoods, we hypothesized that one’s own exposure to immigrants would affect how one responded. Just as earlier studies have documented that Swedes are more pro-immigration, our study confirmed that the Stockholm respondents, despite living in a more segregated city, were clearly more positive towards ethnic mix than are their fellow Nordic capital residents in Helsinki and Oslo. We cannot know whether this has to do with longer or greater exposure to immigrants or whether it has to do with more fundamental historical social factors. At least, it does not seem to relate to one’s own neighbourhood ethnic conditions, as the pro-mix attitude was at a similar level across different immigrant densities in Stockholm. In Oslo on the other hand, respondents living in immigrant-dense neighbourhoods reported more positive attitudes towards ethnic mix compared to those living in lower concentrations. Whether this result is driven by selection of pro-ethnic-mix natives into these areas or with their experience of non-Nordic neighbours is unclear.

We expected the ethnic composition of people’s own neighbourhoods to affect not only their views on whether segregation is a problem, but also whether they wished to see changes in the neighbourhood’s ethnic composition. We expected respondents living in neighbourhoods having a high share of non-Nordic immigrants to be more in favour of reducing the immigrant presence than those living elsewhere, given not least the fact that the public debate identifies immigrant-dense neighbourhoods as problematic, which could influence the native-born residents’ views on the ethnic composition as such. We studied this relationship both descriptively and by employing a logistic regression model. Both approaches showed a very strong but non-linear relationship between immigrant density in one’s own neighbourhood and a preference for a lower share of immigrants. Perhaps somewhat surprisingly, this had nothing to do with real migration behaviour: recent native-born neighbourhood movers did not differ from stayers in their views on neighbourhood diversity in general or their preference for share of immigrants in their own neighbourhood (see also Kaufmann & Harris, 2015). However, a couple of things should be added. Firstly, those wanting a higher share of immigrants dropped in a more or less linear fashion from low to high immigrant-dense areas. This category was small in Helsinki but quite noticeable in Oslo and Stockholm. Secondly, if we combine those who preferred the current level with those who preferred a higher level, that combined category outnumbers those wanting a lower share in all but the highest non-Nordic deciles. So, generalizing to the entire population in the three cities, people did not favour less immigrant presence.

This said, we indeed found a relative tipping point: those living in the upper deciles were much more likely in all three cities to state a preference for a lower share of immigrants in their neighbourhood. After controlling for the respondents’ individual characteristics, the clearest difference is, on the one hand, the difference between Stockholm/Oslo and Helsinki among those residing in low immigrant-dense neighbourhoods (less pro-mix in Helsinki), and on the other hand, between Stockholm/Helsinki and Oslo concerning those residing in high immigrant-dense neighbourhoods (more pro-mix in Oslo). We also underscored above that people seemed to view low and high shares of immigrants in relation to the city context. Despite the fact that high immigrant density represented relatively low concentration in Helsinki compared to Stockholm and Oslo, a high level in the Helsinki context triggered sentiments and opinions regarding immigrant density similar to those in the other cities. This relational dimension is important to acknowledge to better understand popular imaginations and preferences. And this finding certainly needs to be explored further in comparative research on ethnic segregation.

In terms of policy implications, we found fairly strong support for accepting immigration into the cities and for living in ethnically mixed neighbourhoods, but our results also indicate that high levels of immigrant concentration should be avoided, at least if support from native-born residents is to be maintained.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This work was supported by the Norface program on Migration in Europe – Social, Economic, Cultural and Policy Dynamics [project number 399].

Acknowledgements

We would also like to extend sincere thanks to the three reviewers who provided constructive comments on an earlier draft of this paper.