Has multiculturalism failed in Britain?

Abstract

This paper subjects the criticisms advanced against multiculturalism to empirical test. It asks whether ethno-religious groups lead ‘parallel lives’ and, in consequence, fail to integrate with the wider society. It looks in particular at the alleged corrosive effects of multiculturalism, specifically at the maintenance of an ethnic rather than a British identity, social distance from white people and willingness to contemplate violent protest, but finds that all groups alike have displayed major change across the generations in the direction of a British identity and reduced social distance. It finds no evidence that rates of intergenerational change have been slower among ethno-religious groups that have made successful claims for cultural recognition. In contrast, lower levels of integration are associated with perceptions of individual or group discrimination.

Keywords:

• multiculturalism
• generations
• bonding social capital
• social distance
• national identity
• violent protest

Introduction

A widespread belief in western countries is that multiculturalism, defined as a programme for giving recognition to ethno-religious groups and their cultures, has failed and is instead leading to the entrenchment of separate communities with corrosive consequences for trust and solidarity. This argument has been forcibly expressed by politicians, with Prime Minister David Cameron (echoing remarks of German Chancellor Angela Merkel) claiming that:

Under the doctrine of state multiculturalism, we have encouraged different cultures to live separate lives, apart from each other and apart from the mainstream … . We've even tolerated these segregated communities behaving in ways that run completely counter to our values … . This hands-off tolerance has only served to reinforce the sense that not enough is shared. And this all leaves some young Muslims feeling rootless. And the search for something to belong to and something to believe in can lead … [to] … a process of radicalisation. (Cameron 2011)

Similar arguments have been put forward by academics. For example, Sniderman and Hagendoorn (2007, p. 5) have argued that:

Britain and the Netherlands have promoted multiculturalism to expand opportunities for minorities to enjoy a better life and to win a respected place of their own in their new society. It is all the more unfortunate, as our findings will show, that the outcome has been the opposite – to encourage exclusion rather than inclusion.

Similarly Wolfe and Klausen (2000, p. 29) have claimed that ‘if groups within the nation state receive greater recognition, it must follow that conceptions of overarching national solidarity must receive less.’

While there are many different versions of the critique of multiculturalism, there is a central core of arguments on which we focus in this paper. The first step in most critiques such as Cameron's is the claim that multicultural policies (MCPs) will tend to foster separate communities where ethnic groups lead parallel social lives. More sociologically, the hypothesis is that MCPs promote ‘bonding’ rather than ‘bridging’ social capital.

The second step in the argument is that these parallel communities will preserve ethnic norms and values including some (such as the treatment of women, which is a particular focus of many critics) that run counter to the norms of the broader society, will inhibit identification with the broader society, and will hence lead to hostility and distrust between majority and minority. Thus Barry (2001, p. 88) has argued that ‘a situation where groups live in parallel universes is not one well calculated to advance mutual understanding or encourage the cultivation of habits of co-operation or sentiments of trust.’

The third step is that these separate communities and lack of identification with the wider society will provide fertile soil for radicalization. Thus, social segregation was a major theme in the Cantle (2005, p. 9) report on the disturbances (basically riots) in Oldham, Burnley and Bradford, and the report emphasized the way in which social segregation had had adverse implications for conflict and disorder.

A central assumption of many critics is that the problematic consequences of MCPs will apply particularly to Muslims. As we shall see, in Britain some Muslim communities do indeed have high rates of in-group marriage and of course the 7/7 bombings were carried out by young British-born Muslims. However, many of the arguments would seem to apply equally to the case of Sikhs, who have won some notable battles, for example over the right to wear the turban, for cultural recognition.

A second key but implicit assumption is that MCPs will entrench these patterns of parallel lives across generations. New migrants tend to rely on their ethnic communities for social, emotional and practical support, and we might expect this tendency to be greater for groups whose members do not speak English or experience other barriers to social integration. However, critics of multiculturalism are implicitly making a stronger claim than simply that new migrants from certain origins will turn to their ethnic communities for support. They imply that MCPs will entrench this tendency and maintain the strong inward-looking communities across time and generations, thus preserving ‘bright’ boundaries between minorities and the wider society.

Another key assumption is that bonding social capital, and an emphasis on preserving communities' traditional practices, is incompatible with bridging social capital or a commitment to British values. They are seen to be mutually exclusive ‘either/or’ phenomena. Similarly, there is also an implicit assumption that segregation or assimilation are the only possibilities. A helpful framework for analysing this is provided by Berry (1992), who argues that there are two important issues that individuals and groups confront in culturally diverse societies. The first relates to the maintenance of one's own cultural identity and the second relates to the nature of relationships with the wider society. This leads Berry to present a fourfold typology of ‘acculturation options’, namely assimilation, integration, separation and marginalization, which we reproduce in Figure 1.

Figure 1 Four acculturation strategies

Source: Berry1992.

So the critics are essentially concerned that some groups, notably those that have been given the benefit of MCPs, may be entrenched in the bottom-left cell of Berry's diagram (Figure 1), and that radicals and extremists will emerge from groups located in this cell. A counter expectation is that in a liberal society such as Britain, which has implemented a ‘thin’ version of multiculturalism, the normal pattern would be for groups to occupy the top-left cell with positive orientations both to their ethnic culture and to the wider society, and with dual identities rather than exclusively ethnic or British ones.

But even if we were to find that some groups were entrenched in the bottom-left cell, this would not necessarily indicate that MCPs were responsible. An alternative hypothesis is that the wider society's treatment of minorities might drive some groups into this category. Theories of discrimination and reactive ethnicity thus become an alternative explanation for entrenchment and corrosion.

There has been relatively little empirical research on these issues. Some studies such as those of Sniderman and Hagendoorn (2007) and Kesler and Bloemraad (2010) have explored the effects of MCPs on the majority population (or in Kesler and Bloemraad's case, on the population as a whole), reaching very different conclusions from each other. However, our focus is on the effects on the minority population, which has been a particular concern of the critics.

One major study of the effects on the foreign-born population is that of Koopmans (2010), who focuses on three indicators of socio-economic integration (economic activity, residential segregation and over-representation in the prison population) in five European countries. He finds that immigrants to Britain score relatively well on two of these outcomes, but attributes this to Britain's rather lean social security system. He suggests that it is the combination of a generous social security system and MCPs that is particularly harmful to immigrant integration. In another major cross-national study focusing on foreign-born populations, Wright and Bloemraad (2012) look at the effects of MCPs on measures of social inclusion, political inclusion and political integration. They find that countries with stronger MCPs do not have worse outcomes with respect to these aspects of immigrant incorporation. They also find some suggestions that the combination of MCPs with high access to citizenship may have positive effects on some outcomes such as trust, although they admit that the evidence is somewhat equivocal.

While Wright and Bloemraad's concerns come closest to our own, they (like Koopmans) limit themselves to foreign-born populations rather than to the second or later generations. Our contention is that there may be many reasons why the foreign-born, particularly those who migrate as adults, might show some of the characteristics that critics object to. Thus high levels of ethnic, rather than British identity, might be natural among those who may intend to return home or who may not have lived long enough in Britain to be eligible for citizenship. The key test, we believe, is whether such lack of integration is entrenched across generations. Following standard accounts of generational change (e.g. Alba and Nee 1999), we would expect the second and later generations to have higher levels of integration. The key question for us is whether MCPs inhibit this ‘normal’ process of social and political integration across generations.

We focus on three aspects that critics have emphasized: identification with Britain, hostility towards or lack of sympathy with the mainstream, and political radicalization. While a comparative research design would have been ideal, lack of comparative data sets with the requisite second-generation samples makes this very difficult. Instead we focus on a comparison within Britain between groups that have been given more and less cultural recognition. Broadly speaking, ethno-religious groups such as Muslims and Sikhs have asked for, and received, exemptions from standard British rules (such as provisions for halal meat for Muslims or wearing the turban for Sikhs). The 2006 Racial and Religious Hatred Act and the 2006 Equality Act, which outlawed discrimination on grounds of religion or belief, also addressed the concerns of faith groups, particularly following Muslim concerns after the Salman Rushdie affair. In contrast, black groups have tended to make claims for anti-discrimination measures (see Statham 1999, Table 7, p. 617), which, following Barry, we do not treat as multicultural demands¹ (and which also apply across the board and not just to blacks).

The central questions that we address in this paper, therefore, are whether there has been the expected intergenerational shift away from parallel lives and separate cultures and towards the British mainstream, or whether parallel lives and separate cultures have become entrenched across generations among those ethno-religious groups such as Sikhs and Muslims who have received state support for their cultural concerns. Our aim is not to debate the philosophical arguments about the justification of multiculturalism but to assess empirically whether or not the corrosion anticipated by the critics has actually occurred in Britain. We focus on the effects on the minorities rather than on the corrosive effects on the majority group (which would need a different research design).

The plan of the article is as follows. After describing our data and the key variables, we begin with descriptive results, charting differences across generations in the extent to which different ethno-religious minorities lead parallel lives. We then investigate the alleged corrosive effects of parallel lives, looking at changes across generations in value orientations, identification with Britain, social distance from the majority group, and willingness to support violent protest. We then model the data and test whether patterns of generational change differ between ethno-religious groups and whether levels or rates of change are related to experiences or perceptions of exclusion and prejudice.

Data and measures

Our data come from the 2010 Ethnic Minority British Election Survey (EMBES). EMBES provides a representative probability sample of the major established ethnic minorities in Britain – namely people of Indian, Pakistani, Bangladeshi, black Caribbean and black African background. The design was a clustered, stratified sample with areas of high ethnic minority density being over-sampled and areas with the lowest density (< 2 per cent ethnic minorities) being excluded for cost reasons. (In order to adjust for this over-sampling, descriptive statistics are weighted, while in the statistical modelling we control for ethnic density.) Addresses were screened for the presence of ethnic minority individuals. The achieved sample size was 2,787 respondents with a response rate between 58 and 62 per cent, depending on the method of treating ‘unknowns’ in the screening process. The study was funded by the Economic and Social Research Council (ESRC), and the Electoral Commission was a collaborating partner in this research. For further technical details and for the full questionnaire, see Howat et al. (2011).

Two key variables in this paper are ethno-religious group and generation. We next describe how we constructed these variables. Details of the further variables used in the analysis are given in Appendix 1.

Ethno-religious group

We constructed a combined ethno-religious variable since the Indian and black African groups are highly heterogeneous in ways that are relevant to MCPs. We therefore distinguish within the Indian group between those who are Hindu, Sikh, Muslim or belong to other or no religions. This is because Sikhs can legitimately claim to be an ethnic group in their own right and because a great deal of the multiculturalism debate has focused on Muslims.

We also distinguish within the black African group between Muslims and those who belong to other religions – largely Christian. Ideally, if sample size permitted, we would make further distinctions, particularly between Pentecostal, Anglican and Catholic Christians, and between different language groups. There are further ethnic or linguistic differences within our other ethno-religious groups too. For example, within the group of Pakistani background one could distinguish Urdu, Sindhi and Pashto speakers. We should note that the Pakistani and Bangladeshi groups do contain small numbers (around 5 per cent) who are not Muslim, but there are too few to permit a sensible distinction in our data set.

Generation

We have found it helpful to distinguish between a ‘first generation’, who were born abroad and arrived as adults at age sixteen or older, a ‘1.5 generation’ who were also born abroad but migrated before age sixteen, and a ‘second generation +’ who were born in Britain and which also includes the small number of third-generation individuals whose parents were born in Britain. For simplicity, we shall refer to this latter category simply as the second generation.

The broad rationale for making these distinctions is that the first generation, thus defined, will have received most or all of their education in the country of origin and will have been socialized into origin-country values and identities. In contrast, the 1.5 generation will have received some or all of their education in Britain and will have had more exposure to British values in their formative years. Finally, the second generation will have received all their education in Britain and will in almost all cases (99 per cent in our sample) be British citizens. We should note that the black Caribbean group has the highest percentage in the second generation, whereas the black African groups, being the most recent arrivals in Britain, are predominantly first generation. This further means that for some analyses it is not sensible to report percentages for the second-generation black African groups: if the base frequency is less than twenty respondents, we do not report percentages.

Descriptive findings

We begin with the claim that minorities live parallel lives and that this will be entrenched across generations. In Table 1 we show the percentages of our main ethno-religious groups who have a co-ethnic partner, who belong to a voluntary association whose members are mainly co-ethnic, whose friends are mainly co-ethnic, who attend a place of worship comprised mainly of co-ethnic worshippers, whose neighbours are mainly co-ethnic, and who work at a firm where colleagues are mainly co-ethnic. In each column we show the percentage for the first generation followed by the change between the first and second generations. Changes that are statistically significant at the .05 level are highlighted. (For simplicity we exclude the 1.5 generation, although they are included in our modelling later in the paper.)

Table 1 Generational change in social life (percentages; 1st generation on the left and 1^st–2nd generation change on the right of each column)

	Co-ethnic partner		Association members mainly co-ethnic		Friends mainly co-ethnic		Fellow worshippers mainly co-ethnic		Neighbours mainly co-ethnic		Workmates mainly co-ethnic
Bangladeshi	100	−11*	87	−10	62	−13	52	+6	26	+12*	37	−18
Pakistani	96	+2	74	0	61	−6	71	+13*	35	+4	40	−23*
Indian Sikh	95	−2	96	−14*	65	−22*	91	+4	15	0	36	−15*
Indian Hindu	96	0	82	+7	60	−18*	82	+9	16	−6	16	−6
Indian Muslim	88	−11	–	–	75	−24*	62	+3	17	+20	–	–
Black African Muslim	92	–	64	–	53	–	26	–	13	−1	18	−10
Other black African	88	–	68	+4	56	−28*	43	+8	7	−2	18	−15*
Black Caribbean	81	−29*	70	−3	53	−14*	36	−3	10	−3	8	−1
All minorities	92	−11*	77	−3	57	−15*	52	+6*	18	+2	24	−12*

* p<.05 (weighted)

Note: Percentages not reported where n<20. The base excludes non-partnered (column 1), non-members of associations or religions (columns 2 and 4) and non-workers (column 6).

We have arranged Table 1 according to the strength of the ‘bonding’ tendency in the first generation. Thus, we see the highest levels of in-group choice occurring with respect to partnership, where around 90 per cent of the married/cohabiting members of the first generation partner co-ethnics. (Many of these marriages will have been formed in the country of origin before migration.) Associational membership tends to be predominantly with co-ethnics in the first generation and there are also high levels of in-group friendship and of attendance at predominantly co-ethnic places of worship. In-group choice is much lower for residential neighbourhoods and workplaces. We have checked the self-reported figures for residential segregation against official (2001) census figures of the ethnic composition of the areas where our first-generation respondents lived. The figures tally remarkably well.²

It is also striking that the south Asian groups generally show greater co-ethnic choice with respect to partners, friends, church and associational involvement than do the black groups. This applies to Sikhs and Hindus as well as to Muslim Indians, Pakistanis and Bangladeshi, while black African Muslims show lower levels of co-ethnic choice.

As to entrenchment across the generations, the pattern is not straightforward. There is a modest increase in the extent of attendance at predominantly co-ethnic places of worship.³ We see little change with respect to participation in ethnic associations or residential concentration. In contrast there is greater mixing across the generations in friendship and at work, and there are also significant increases in intermarriage. Furthermore, despite different starting levels in the first generation, the direction of change is broadly similar in all ethno-religious groups.

In Table 2 we then turn to minorities' value orientations and attitudes. We asked two questions, designed to tap the two dimensions of Berry's typology of acculturation strategies, on the desirability of maintaining ethnic values and beliefs, and of mixing and integrating. We also included a question, asked only of Muslims, on the desirability of having sharia courts (an indicator of a desire for separatism) and another on support for traditional gender roles.

Table 2 Generational change in attitudes and orientations (percentages; 1st generation on the left and 1st–2nd generation change on the right of each column)

	Minorities should maintain own values		Minorities should mix and integrate		Sharia law should apply to all cases		Husband should work and wife stay at home
Bangladeshi	85	−2	93	−5	19	+3	16	+2
Pakistani	88	−7	92	+1	19	+3	38	−21*
Indian Sikh	92	0	99	−4	N/A		15	0
Indian Hindu	87	+5	93	+6*	N/A		19	−6
Indian Muslim	7	+7	90	0	21	−16*	–	–
Black African Muslim	80	–	91	–	16	–	32	–
Other black African	74	+12	91	−3	N/A		21	–
Black Caribbean	82	+1	97	−4	N/A		26	−14
All minorities	83	+1	93	0	19	0	26	−11*

* p <.05 (weighted)

Note: Percentages not reported where n<20. Gender role question asked in mail-back supplement; sharia question asked only of Muslims.

The most striking aspect of Table 2 is that minorities overwhelmingly support maintenance of their own ethnic customs and traditions alongside equally striking support for mixing and integrating. That is, minorities have positive orientations to the cultures of both origin and destination countries (thus occupying the integrationist cell of Berry's typology). This is equally true of all groups, south Asian and black, Muslim and Christian alike. Furthermore there is no generational change here to speak of. Most members of all ethnic groups, both in the first and second generations, have positive orientations towards both minority and majority cultures. Only around 5 per cent occupy the bottom-left separationist cell of Berry's typology and these are outnumbered by 10 per cent in the top-right assimilationist cell.

Similarly, support for separatism as indexed by the question on sharia courts, amounts to just one fifth of Muslims. To be sure, critics of multiculturalism might be worried that the figure is as high as one fifth, and it would be wrong to deny that some support for separatism does exist – just as we could not deny that there is some white British support for ethnic exclusionism.

We also find some support for traditional gender roles, although even in the first generation only a minority of respondents favour the patriarchal model. Moreover, for the two groups where we have large second-generation samples – Pakistanis and black Caribbeans – there is clear evidence of convergence with the white British figure (14 per cent agreement with the patriarchal statement).

So this is a very different picture from Table 1 on social relationships. Whereas there we saw high levels of bonding social capital and social involvement with co-ethnics, in Table 2 we see positive orientations towards mixing and integrating, and low levels of support for separatism. One of the key assumptions of the critics of multiculturalism, namely that high levels of bonding social capital would lead to unwillingness to integrate, does not therefore appear to be well-founded.

We next turn to feelings of British or ethnic identity, hostility towards integration with whites (as indexed by a classic ‘social distance’ question on attitudes towards intermarriage) and finally, a measure of willingness to engage in a violent protest ‘if the British government was about to start a war that you didn't agree with’. These are three key outcomes and represent our best indicators of whether corrosion has in fact been occurring.

In the case of British identity and social distance from white people we see major changes across generations in the direction of greater integration. Thus we see high levels of British identity even in the first generation, and these levels are higher still in the second generation, with significant increases among Pakistanis and the three Indian groups. We also see much reduced hostility to marriage with a white person in the second generation, although there is no decline, and indeed some increase overall, in willingness to support violent protest. However, as Sanders and his colleagues show in their paper in this special issue, the overall level of support for violent protest in the second generation is no different from that of white respondents.

In summary:

• High levels of bonding social capital coexist with positive orientations towards integration, high levels of British identity and low levels of hostility to white people.

• Intergenerational change is faster for some measures than for others – it is fastest for British identity, slowest for bonding social capital.

• The overwhelming majority of all ethno-religious groups show positive orientations both towards their own ethnic culture and towards integration into British society. Only a small minority take a separatist position, rejecting a British identity, supporting sharia law or supporting violence.

Modelling the data

One possibility is that some of the apparent intergenerational differences observed in Tables 1, 2 and 3 are due to confounding factors such as the differing age distributions of the generations. For example, young men may be more inclined to support violent protests, and so the apparent increase in the second generation may simply be a consequence of the younger profile of this generation. Second, we need to check whether the rates of intergenerational change in our key outcome measures are consistent across the different ethnic groups: do groups such as Sikhs and Muslims show a slower rate of generational change than other groups who have not asked for or received MCPs? Third, is the potential support for violent protest observed in Table 3 related to experiences of discrimination and a sense of social injustice rather than to multiculturalism?

Table 3 Generational change in identity, social distance and support for violent protest (percentages; 1st generation on the left and 1^st–2nd generation change on the right of each column)

	Feel equally or more British than black/Asian		Bothered about marriage with white person		Might support violent protest against war
Pakistani	67	+18*	30	−17*	6	+13*
Bangladeshi	63	+12	20	−7	8	+10*
Indian Sikh	56	+30*	25	−17*	14	−2
Indian Hindu	54	+18*	18	−13*	6	−1
Indian Muslim	63	+32*	18	−16	13	−8
Black African Muslim	55	–	11	–	3	–
Other black African	48	+11	14	−2	8	+10*
Black Caribbean	51	+8	10	−4	6	+1
All minorities	57	+22*	18	−10*	7	+5*

* p <.05 (weighted)

Note: Percentages not reported where n<20.

We focus on three outcomes: preferring a black or Asian identity to a British one, social distance from white people, and support for violent protest. For each outcome we fit three models. The first includes demographics (age, educational level, gender, generation and marital status) as well as our ethno-religious group measure. (We also experimented with numerous other potential confounding variables such as citizenship, fluency in English, income source and social class but none of these proved to be consistent predictors, although speaking a language other than English at home was positively related to having an ethnic identity.) We control throughout for percentage co-ethnics in the Lower Super Output Area (LSOA) in order to take account of the over-sampling in high-density areas.

In the second model we test formally whether rates of intergenerational change differ across the nine ethno-religious groups. In the third model we then include various mediating variables, focusing on the key variables identified in Table 1, together with measures of personal experience of discrimination and feelings that their ethnic group is held back by discrimination. We need to recognize, however, that we cannot be sure in cross-sectional research whether the ‘mediating’ variables genuinely have a causal role: it is always possible that, for example, an ethnic identity leads one to join an ethnic association, rather than the other way round. In order to gauge the substantive magnitude of these associations, in Table 4 we show the average marginal effects, which can be interpreted straightforwardly as the change in the probability of the outcome occurring.

Table 4 Logistic regression of identity, social distance and support for violent protest (average marginal effects)

	Feels more black/Asian than British		Bothered about marriage to white person		Might support violent protest against war
	Model 1	Model 3	Model 1	Model 3	Model 1	Model 3
Ethno-religious group
Pakistani (ref)
Bangladeshi	0.07*	0.06	−0.04	−0.04	0.01	0.00
Indian Sikh	0.08*	0.05	−0.07**	−0.08*	0.04	0.02
Indian Hindu	0.11**	0.10**	−0.09**	−0.08**	−0.01	−0.02
Indian Muslim	−0.02	−0.01	−0.09*	−0.08	−0.02	−0.02
Other Indian	0.05	0.03	−0.12**	−0.13**	0.00	0.00
Black African Muslim	0.13***	0.14**	−0.10*	−0.11**	−0.02	−0.03
Other black African	0.25***	0.20***	−0.03	−0.05	0.04**	0.02
Black Caribbean	0.24***	0.21***	−0.07**	−0.09***	−0.03*	−0.04
Generation
1st (ref)
1.5	−0.13***	−0.14***	−0.06**	−0.07**	0.06***	0.05***
2nd +	−0.25***	−0.27***	−0.11***	−0.12***	0.06***	0.05***
Age (ln)	−0.31***	−0.33***	−0.04	−0.04	−0.08**	−0.09**
Gender	0.01	0.01	0.01	0.02	−0.03**	−0.03**
Married	0.00	0.00	0.03*	0.03*	0.00	0.00
Education	0.00	0.00	−0.01*	−0.01**	0.00	0.00
Interactions between generation and ethno-religious group	NS		NS		NS
Co-ethnic friends		0.03**		−0.02		0.00
Co-ethnic neighbours		0.01		0.01		−0.01
Co-ethnic worship		−0.02		0.01		0.02
Co-ethnic association		0.08***		0.06***		0.05**
Group discrimination		0.25***		0.06		0.10***
Individual discrimination		0.09***		0.07***		0.00
Percentage co-ethnic (ln)	0.06	0.04	0.11**	0.09*	−0.01	0.00
Index of multiple deprivation (ln)		0.08		−0.03		−0.04
N	2,552	2,552	2,537	2,537	2,520	2,520

* p<.05, ** p<.01, *** p<.001 (unweighted)

Note: Models run in STATA with robust standard errors. The average marginal effects (AME) report the change in the dependent variable computed at different values of the x variables and then averaged across the different values. For dummy variables, the marginal effect shows by how much the probability of the outcome is predicted to change as x changes from 0 to 1 holding all other explanatory variables equal. For ordinal variables, the AME report the predicted change for an increase in x of one unit. For the continuous variables, we report the difference between the probability of the outcome occurring for the lowest and highest value that the variable can take.

Considering identity first, we can see from model 1 that a black or Asian identity is negatively associated with age (i.e. younger people are more likely to adopt a black or Asian identity). After controlling for age, we find that the second generation is markedly more likely to feel British than the first generation (up by 25 percentage points), with the 1.5 generation in between. It is also notable that it is not the Muslim groups but the black groups, especially those of black Caribbean heritage, who are least likely to feel British and most likely to adopt a black identity. In model 2 we then test whether rates of generational change vary across ethno-religious groups (i.e. we tested for interaction effects). The improvement in fit when these interactions were introduced was not significant (overall change in χ²=11.93 with 8 d.f., p>.20), while only one of the individual terms was significant – that for other black Africans, who showed a slower rate of intergenerational change.

In model 3 we then introduce mediating variables to see how far ethno-religious or generational differences can be explained by differences in bonding social capital or experiences of discrimination. Our measures of bonding social capital are co-ethnic friendship, neighbours, places of worship and ethnic associations. Of these variables, membership of a co-ethnic association has the strongest association with a black or Asian identity.

We also include measures of reported individual and group discrimination. As we can see, both forms of discrimination increase one's likelihood of feeling black or Asian rather than British, with perceptions that the group is held back by discrimination proving to be more important than any of our other predictors except age and generation (a difference of 25 percentage points between those who are most aware and those least aware of group discrimination).

Turning next to social distance, we again see considerable generational progress. However, the ethno-religious differences are almost the exact opposite of those found for identity. In the case of social distance, we find that the black groups are much less bothered about intermarriage with a white person than are the south Asian groups, with people of Pakistani background showing the highest level of concern. In other words, despite their lesser acceptance of a British identity, black groups show a greater acceptance of white people.

As with identity, the introduction of interaction terms between ethno-religious group and generation does not improve fit (change in χ²=10.31 with 8 d.f., p>.25), and again it is the other black Africans who show the slowest rate of intergenerational change. Our mediating variables, too, operate in much the same way as with identity; model 3 shows that membership of a co-ethnic association and perceptions of individual and group discrimination are associated with greater social distance from white people”.

Finally we turn to support for violent protest. Here we get a different pattern. The generational trend, albeit modest in size (a change of only 5 percentage points between the first and second generations), is towards greater support for protest. Furthermore, unlike identity and social distance, there are no significant ethno-religious differences in support for violent protest: in particular, Muslims show no greater support for violence than do other south Asians or blacks. Nor is there any sign that rates of generational change differed between ethno-religious groups. The introduction of the interaction terms once again did not improve fit (change in χ²=7.84 with 8 d.f., p>.25) and none of the individual terms approached significance.

However, support for violent protest is associated with perceptions of group discrimination (a 10 percentage-point difference), as is membership of ethnic associations (5 percentage points). However, both these ‘effects’ are substantively rather small and we should therefore be careful not to exaggerate the importance of ethnic organizations or generational change as a source of violent protest. The other distinguishing characteristics of people who might be prepared to engage in violent protest are that they are young and male. This profile almost certainly fits white British protesters too and probably has nothing to do with Muslim or ethnic radicalization.

Conclusions

The first lesson to be drawn from our evidence is that bonding social capital does not have the adverse consequences anticipated by the critics. On the one hand, it is true that some south Asian groups, particularly those of Pakistani and Bangladeshi background, do exhibit high levels of in-group marriage and friendship, but they do not lead parallel lives since residential and workplace segregation is actually rather low. We also find that people of Pakistani and Bangladeshi background are the most likely to feel bothered about intermarriage but on the other hand they are no more inclined to reject integration into British society, to reject a British identity, or to contemplate violent protest than are other ethno-religious groups. Indeed, it is the black groups who are the most likely to reject a British identity, while there are no significant ethnic differences in the propensity to contemplate violent protest. In short, high levels of in-group marriage and friendship are compatible with the adoption of a British identity and a positive orientation towards British society. As Table 4 showed, the only one of the ‘bonding’ measures that consistently predicts our three outcomes is membership of an ethnic organization.

The second lesson to be drawn is that intergenerational changes do not fall into a neat pattern with Muslim and Sikh groups being more entrenched than the black groups. If anything, it was the other black African group that showed the slowest rate of intergenerational change. There was no sign in our data that the groups that had made successful claims for recognition of their cultural practices, such as Sikhs and Muslims, had made slower intergenerational progress towards integration. The implication is that MCPs have played little part in the story.

A third lesson is that perceived discrimination (both individual and group) has some of the strongest effects on negative outcomes. Discrimination is at least as plausible an explanation as multiculturalism for lack of integration. In this respect our results are consistent with those found by Maxwell (2009), who has shown (using a quite different data set) the importance of perceived discrimination for lack of British identification.

Overall, our results paint a rather optimistic picture of ethnic minority integration in Britain. In this respect, our results are consistent with those of Koopmans (2010) and of Wright and Bloemraad (2012), both of whom show that migrants in Britain compare relatively favourably with those in other countries on various measures of social and political integration. Like Wright and Bloemraad, we have found no evidence that MCPs have had negative effects on social integration: the similarity of rates of intergenerational change for the different ethno-religious groups, albeit from rather different starting points, suggests that we are seeing in Britain general processes of intergenerational integration that have little to do with specific MCPs. As Wright and Bloemraad (2012, p. 88) argue:

Academic research tends to downplay null findings, but in this case these findings carry enormous theoretical and policy significance: the most important rationale for the political backlash against multicultural policies – that they impede or hurt socio-political integration – appears unfounded empirically.

We are in complete agreement.

Acknowledgements

We gratefully acknowledge the ESRC, which funded the EMBES study, and the Electoral Commission, which was a collaborating partner in our project. We would also like to thank our colleagues on the EMBES study, Stephen Fisher, Gemma Rosenblatt, David Sanders and Maria Sobolewska, and our colleagues at TNS-BMRB, Nick Howat, Oliver Nordern, Emily Pickering and Joel Williams, who made such a success of the fieldwork.

Additional information

Notes on contributors

Anthony Heath

ANTHONY HEATH is Professor of Sociology at the University of Manchester and Emeritus Professor in the Department of Sociology at Oxford University

Neli Demireva

NELI DEMIREVA is Postdoctoral Research Fellow in the Department of Sociology at Oxford University

Notes

1. While Banting and Kymlicka (2004) include affirmative action policies as one of the components of their index of MCPs, we do not regard anti-discrimination policies as falling within the definition of MCPs since they do not focus on cultural rights but aim to ensure simply that minorities achieve the same opportunities as do the majority population.

2. According to the 2001 census, the co-ethnic proportion in the LSOAs where our first-generation Caribbeans lived was 9 per cent. The corresponding figures were 14 per cent for Bangladeshis, 22 per cent for Sikhs and 23 per cent for Pakistanis.

3. This may be because the size of the different ethno-religious communities has grown and it has thus been easier for the second generation to attend an ethnic mosque rather than a mixed one. It is also possible that the second generation have a different interpretation of our question that refers to ‘ethnic background’.

Appendix 1. Definitions of variables used in Table 4

Independent variable
Age	Natural logarithm of age
Gender	0 = male; 1 = female
Married	0 = not married; 1 = married/cohabiting
Education	0 = none; 1 = low; 2 = GCSE; 3 = A-level; 4 = degree
Generation	0 = 1st generation; 1 = 1.5 generation; 2 = 2nd generation plus
Ethno-religious group	0 = Pakistani; 1 = Bangladeshi; 2 = Indian Sikh; 3 = Indian Hindu; 4 = Indian Muslim; 5 = other Indian; 6 = black African Muslim; 7 = other black African; 8 = black Caribbean
Co-ethnic friends	‘As far as you know, how many of your friends have the same ethnic background as you?’ 5 = all of them; 4 = most of them; 3 = about half of them / don't know; 2 = a few of them; 1 = none of them
Co-ethnic neighbours	‘As far as you know, how many of the people in your neighbourhood have the same ethnic background as you?’ 5 = all of them; 4 = most of them; 3 = about half of them / don't know; 2 = a few of them; 1 = none of them
Co-ethnic association	0 = does not participate in any association or participates in non-ethnic association; 1 = participates in ethnic association
Co-ethnic worship	0 = not a church member or non-ethnic church member; 1 = member of co-ethnic church
Individual discrimination	0 = not experienced race discrimination; 1 = experienced race discrimination
Group discrimination	‘There is a big gap between what people from my ethnic group expect to get out of life and what they receive.’ ‘Non-white people don't have the same opportunities and chances in life as white people, as they are held back by prejudice and discrimination.’ Both items recoded so that 1 = strongly disagree; 2 = disagree; 3 = don't know / neither agree nor disagree; 4 = agree; 5 = strongly agree. Scores summed.
Percentage co-ethnic	Natural logarithm of percentage co-ethnics in the LSOA
Index of multiple deprivation	Natural logarithm of the index of multiple deprivation
Dependent variable
Social distance	How bothered if a close family member married a white person: 0 = not at all / not very much / a little / no opinion; 1 = rather a lot / a great deal
Feel more black/Asian than British	0 = equally (black/Asian) and British / more British than (black/Asian) / British not (black/Asian) / other answers; 1=(black/Asian) not British / more (black/Asian) than British
Might support violent protest against war	0 = no / don't know; 1 = yes