Early View
Open Access

A tale of four cities: Neighbourhood diversification and residential desegregation in and around England's ‘no majority’ cities

Richard Harris

Corresponding Author

Richard Harris

School of Geographical Sciences, University of Bristol, Bristol, UK


Richard Harris, School of Geographical Sciences, University of Bristol, University Road, Bristol BS8 1SS, UK.

Email: [email protected]

Search for more papers by this author
First published: 15 November 2023


The publication of the 2021 Census data revealed that four English cities—Birmingham, Leicester, London and Manchester—are now ‘no majority’ cities, meaning that no ethnic group, including the White British, comprise a majority (more than half) of their populations. This paper explores the residential diversification of these cities to ask: whether that diversification is reflected in the average neighbourhood of all ethnic groups or just some; whether the decline in the number of White British means that ‘enclaves’ of other ethnic groups are emerging instead; whether the White British are avoiding living in diverse neighbourhoods; and whether a co-occurrence of the diversification is that residential segregation between the White British and other groups is increasing within and beyond the boundaries of these cities. Using a harmonised set of cross-census neighbourhoods to provide a consistent geography across the 2001, 2011 and 2021 censuses, the results show that the residential neighbourhoods of the four cities have increased their ethnic diversity for the average member of all ethnic groups. Despite some growth in the number of neighbourhoods where a group other than the White British form a majority, especially in Leicester, the overall conclusion is one of residential diversification happening alongside residential desegregation.

Short Abstract

Birmingham, Leicester, London and Manchester have emerged as England's only ‘no majority’ cities, where no single ethnic group forms more than 50% of the population. This paper looks at where the diversification of these cities has continued over the period from 2001 to 2021 or whether the reduction in the number of White British has led to their replacement by another dominant group in some neighbourhoods. It also looks at whether diversification is associated with segregation of the White British from other groups within or beyond the boundaries of these cities and finds that it is not.


This paper offers a detailed study of ethnic diversification and desegregation in and around the neighbourhoods of England's only four cities in which no single ethnic group forms a majority (more than half of the population). These are Birmingham, Leicester, London and Manchester. Their lack of a majority group suggests that they are ethnically diverse cities. However, it is possible that one previously dominant group has been replaced by another, albeit one that does not form a majority. It is also possible that diversification within these cities is driving segregation within or beyond them. With that in mind, this paper concentrates on answering four questions. First, is the ethnic diversity of these four cities experienced by all ethnic groups within them or do some groups live in more diverse neighbourhoods, on average, than others? Second, does the decline in the number of White British residents mean these cities now contain fewer neighbourhoods that are predominantly or majority White British and are ‘enclaves’ emerging of other ethnic groups instead? Third, did the number of White British decline fastest in neighbourhoods that had already experienced greatest diversification? Fourth, if the White British are leaving these cities for surrounding and often less urban areas, are the areas they appear to be moving to also diversifying or is the process resulting in greater segregation between the White British and other ethnic groups?

Interest in these questions is framed by earlier work on the topic of ethnic de-/segregation and diversification in England. Following the publication of the 2011 Census data, a number of scholars turned their attention to whether the growing diversity of the population—a demographic shift with origins both in historic immigration from parts of the former British Empire and, at that time, in ‘free movement’ from the European Union—was reflected in the residential patterns of the various ethnic groups now living in the country. Greater ethno-cultural diversity has the potential to diversify neighbourhoods and communities, especially within the towns and cities to which immigrants have been attracted, but it also creates the possibility of ethnic residential segregation, whereby different ethnic groups live in separate places from one another.

The publication of the 2011 small area census data allowed the residential patterns of various ethnic groups to be studied and compared with the 2001 data to provide empirical evidence for or against contemporaneous concerns about segregation and social polarisation (Cantle, 2006). Pooling together the various studies, Harris and Johnston (2020) note a dual process whereby ‘minority’ ethnic groups have spread out from locations where they were previously concentrated, thereby contributing to a greater population mix within neighbourhoods, but with spatial contraction of the White British away from cities such as London. This dual process is generally one of diversification and desegregation, but it also led to claims of ‘white flight’ and ‘white avoidance’ (Kaufmann, 2013, 2016) whereby the White British select less diverse neighbourhoods than other groups.

Harris and Johnston observe an asymmetry in the political debate about segregation in the country, including in recent government-sponsored publications that focus on the generally rare examples of neighbourhoods or schools being dominated by a single, non-white group but not on the far more common examples of where the White British are dominant. A consequence of this asymmetry is to allow the populist media, as well as populist political parties, to simultaneously blame immigration for segregation while presenting the White British—and, by extension, ‘traditional British values’—as under threat: it taps into the identity politics and potential (but disputed) response of the majority group to demographic change explored by Kaufmann (2018). That narrative of marginalisation rose again following the publication of the 2021 Census data, where some commentary focused on the White British population being a ‘minority’ in some English cities. The Telegraph newspaper, a right-wing broadsheet, carried an article headlined ‘White British people now minority in UK's two largest cities [London and Birmingham], census reveals’ (Swerling, 2022). A near identical headline appeared on the website of LBC, a national ‘talk’ radio station (Heren, 2022).

Of course, it could be argued that these are only passing headlines and should be treated accordingly. However, the politicisation of processes of demographic change is ongoing and so it is important to show that such claims and their racialised undertones are deceptive. It is true that the 2021 Census figures reveal 37% of London's residential population identify as White British. In Birmingham, it is 43%. The headlines might also have referenced Leicester (33%) and Manchester (but only just; 49%). But these are four of 55 English cities and, for three of them, the White British remains by far the largest group. Only in Leicester is the Indian population fractionally larger than the White British, although both comprise about one third of its population. Describing what is typically the largest group as a minority is a misleading use of language. In none of Birmingham, Leicester, London or Manchester are the White British really a minority in the way that other smaller groups are. While these English cities have been described as ‘majority-minority’ in the sense that the majority of their populations are from ‘minority’ (i.e., not White British) groups, the shorthand hides more than it reveals. A better epithet would be no majority cities.

This paper develops a harmonised set of cross-census neighbourhoods and data to answer the research questions set out at the beginning of this introduction. This offers geographical consistency by retaining the same boundaries for each of the 2001, 2011 and 2021 censuses. Using these data, the study finds no evidence of rising ethnic residential segregation within or extending beyond the boundaries of the four cities. Instead, residential diversification is happening alongside residential desegregation. The results complement Catney et al.'s (2023) ‘rapid response analysis’ of the 2021 Census, but place greater focus on what is happening in and around the four specific cities of interest.


The study of residential ethnic segregation is well established over many decades of research. Much of the literature originates in the United States (e.g., Clemence, 1967; Cowgill, 1956; Lieberson, 1961; Rice, 1968; Taeuber, 1968), but studies of segregation span worldwide (see, inter alia, Anderson, 2020; Nightingale, 2012; Pryce et al., 2021; Schnell & Ostendorf, 2002; van Ham et al., 2021).

The motivations for these studies are multi-faceted and include understanding how processes of demographic change, including historic or present immigration, or the social treatment of non-hegemonic groups, combine with and reinforce processes of socio-spatial inequality—including unequal educational opportunities, housing markets, employment structures and discrimination—to create and recreate enclaves of particular ethno-cultural groups, evidenced by a residential separation from other groups. The (re-/)production of these residential geographies connects spatial segregation with ethnic and social (class) inequalities, where the ethnic and the social are interlinked (Ezcurra & Rodríguez-Pose, 2017; Harris et al., 2017; Massey, 1981; Massey et al., 2009) although not reducible to one or the other (Järv et al., 2020). Residential and spatial clustering may also arise as a result of homophily, the preference to live near people with a shared culture, religion, ethnicity and/or other identity, and it may bring benefits to them (Bakens & Pryce, 2019; Krysan & Crowder, 2017; Merry, 2013). Whether it is truly an unconstrained choice is a moot point. Living close to those who share the same traditions and language as yourself might be a necessary and socially coerced bulwark against the sorts of everyday exclusions and structural prejudices encountered as a member of a minority group or because of other financial or employment constraints that limit residential choices.

Concerns around segregation can be rooted in contact theory and the presumption that living apart undermines cross-community cohesion, tolerance, understanding and the opportunities for shared social capital (Laurence et al., 2019; Sturgis et al., 2014), especially where the separations overspill into other areas of life. Political interest in segregation is sometimes framed within a broader citizenship agenda and fears about a lack of social integration (Cantle, 2001; Casey, 2016). Such fears often place expectations on the minority groups to integrate in terms set by the political classes, rather than reflecting more fully on the structural barriers that prevent it from happening.

Occasionally, a discourse of ‘ghettos’ arises from the realisation that there are some (but few) places in the UK where a high percentage of the population is of a particular ethnic group, although the term is never applied to the White British that is most likely to be living in mono-ethnic communities and attending the least diverse schools (Harris & Johnston, 2020). ‘Ghettos’ of any other ethno-religious group—it is usually the presumed segregation of British Muslims that hits the headlines (Tahar et al., 2023)—is largely a hyperbolic imagining of the more alarmist sections of the British media and some social commentators (see Phillips [2006] for a challenge to discourses of Muslim self-segregation, and Johnston et al. [2016b], for evidence that most Muslims live in neighbourhoods where they are in the minority). However, the word does appear alongside a typology of neighbourhood types developed by Johnston et al. (2002) before being dropped in later papers. Even in that paper, the point was less that ethnic ghettos exist in English cities and more that they are exceptionally rare, lacking comparability with the racial divisions that can be found, for example, in some US cities. As the paper states, ‘there is little evidence of much ethnic group concentration into polarised enclaves and ghettos in English cities, save for a small number of groups in a few places’ (p. 601).

That was in 2001. Although two recent government-backed reports have propagated the idea that ethnic segregation has grown, in some places or among some groups at least (Casey, 2016; HM Government, 2018) the bulk of empirical evidence showed the opposite between the 2001 and 2011 censuses: residential neighbourhoods have diversified, ethnic segregation has fallen, fewer (including the White British) are living in almost mono-ethnic neighbourhoods where their own ethnic group dominates (Catney, 2016a, 2016b, 2017; Johnston et al., 2013, 2015; Johnston et al., 2016c). It would be surprising if that trend subsequently reversed. Harris and Johnston (2020), Lan et al. (2020) and Catney et al. (2023) suggest it has not.

There is a potential caveat. It is evident in a briefing paper of the 2011 Census data that asks whether neighbourhood segregation has decreased from 2001 to 2011 for large urban districts (Catney, 2013). The answer is largely yes but not for the White British living in Inner London, Outer London, Metropolitan areas or other large cities (combined). The measure of segregation for the White British groups increased by an amount much smaller than the decrease for other ethnic groups, but there was still an increase in 86% of local authority districts in England and Wales.

From the study of the 2011 Census data and its comparison to 2001, two perspectives on the demographic and geographic changes emerged, as discussed in the Introduction to this paper. The first focused on a process of dispersion and spatial diffusion of minority groups across cities and the evidence that ethnic minority groups are less segregated from other groups than in the past. This links to studies of ethnic diversity and neighbourhood diversification, a focus less prevalent in the UK literature than that directly on segregation but with various studies evidencing continuing or increased ethnic diversity both in places that were and were not especially diverse in the past (see, for example, Catney, 2016a; Rees & Butt, 2004; Stillwell & Phillips, 2006). The implication of this is that minority groups do not consistently ‘self-segregate’.

The second observed that the places the minority groups are moving to tend to have declining numbers of White British residents, with a pattern of spatial retrenchment and concentration of the White British out of London and other former industrial and manufacturing cities. Kaufmann (2016) adopted a language of ‘white avoidance’ (see also Andersen, 2017), switching from the language of ‘white attraction’ that appeared in Kaufmann (2013), and has also used the phrase ‘white flight’ (Kaufmann, 2023). Kaufmann (2018) argues that ‘the overwhelming story that the statistical models tell, is one in which whites are moving towards the most heavily white neighbourhoods’ (p. 401). The basis for that claim is the empirical observation that decreases in the number of White British residents tended to be greater between 2001 and 2011 in the local authorities where the White British were already less prevalent and other groups more so (see also, Kaufmann, 2023). This trend may suggest an aversion to changing neighbourhoods or, more bluntly, to other ethno-cultural groups.

Catney (2016a) contests this: ‘The increased share of all ethnic groups (White British and minority) in less urban areas challenges claims of “White flight” from diversity’ (p. 750). Simpson and Finney (2009) show that counter-urbanisation is common to all ethnic groups except Chinese, stating ‘“White flight” is not an appropriate term to describe White movement, nor to explain the growth of ethnically diverse urban areas’ (p. 37). When Harris and Johnston (2020) looked at neighbourhoods with census or school populations that had relatively low percentages of the White British at one time period and reviewed what happened by a later period, they found plenty of occasions where the number of White British grew. Evidence for ‘white avoidance’ was weak, at best. In any case, the simple, bivariate relationship considers neither differential age structures (Sabater & Catney, 2018; Simpson & Jivraj, 2015) nor unequally distributed ‘pull factors’ such as retirement, income or the type of jobs that enable movement to suburban and rural areas. Whist Kaufmann (2023) evidences that the White British select significantly less diverse neighbourhoods than other groups, even after controlling for age, education and income, he also states, with a degree of hyperbole at the beginning of the sentence, that ‘the dramatic aggregate-level ethnic sorting that took place in Britain in the 2000s cannot … be attributed to ideologically motivated white avoidance’ (p. 769).

Nevertheless, the idea of white avoidance has two (unintended?) virtues. First, it reminds us that the residential decisions of the White British have the capacity to drive segregation and potentially more so than the actions of smaller minority groups. Second, it recognises that cities are not closed systems with impermeable boundaries preventing population movement outside of them. Processes of desegregation and segregation can occur concurrently; views of these depend upon the spatial extent and scale by which you look at them (Harris, 2017). There have been various studies of the patterns of residential desegregation in London, for example (Johnston et al., 2016a; Jones et al., 2015), but it is entirely possible for the White British to become more segregated from other groups by leaving London, while those who remain become less segregated from others of London's population. Truncating at the boundaries of cities conceals what is happening across them. For that reason, the analysis that follows proceeds in two broad stages. First, it establishes that the four no majority cities are becoming more residentially diverse and that this is the typical experience of all ethnic groups who live in them, including the White British. It then looks at whether the diversification is occurring alongside greater segregation between the White British and other groups over wider regional and national spaces. It finds that it is not.


3.1 Creating a consistent, cross-census geography

Neighbourhoods used in this study are based on Output Areas (OAs), the smallest geographical areas of data release for each English and Welsh census, containing, in 2021, a mean average of 335 persons in the four cities, with an interquartile range from 273 to 387. The OAs provide a broadly consistent small area geography to study population changes from 2001 onwards, whereas previous censuses had substantial changes from one census to the next. Of 175,434 OAs in England and Wales in 2001, 166,393 (95%) remained unchanged in 2021, which is also 88% of the 188,880 OAs in 2021.

Some, however, are modified, due to new housing developments, for example. To address this, for this study the census OAs for 2001, 2011 and 2021 have been harmonised into a set of cross-census neighbourhoods that are the same in each time period. The method to achieve this is straightforward and enabled by the look-up files, provided at the Office for National Statistics' Open Geography Portal,1 recording whether an OA is changed or unchanged between censuses.

Consider that OA x in 2011 is split into OAs y and z in 2022: x t 11 y , z t 21 $$ {x}_{(t11)}\leftrightarrow {\left\{y,z\right\}}_{(t21)} $$ . We can regard x $$ x $$ , y $$ y $$ and z $$ z $$ as connected where x $$ x $$ is the parent of y $$ y $$ and z $$ z $$ . A consistent geography is created by aggregating y $$ y $$ and z $$ z $$ back into x $$ x $$ and tallying their data, x = y z $$ x=y\cup z $$ . Now suppose that OAs v $$ v $$ and w $$ w $$ from 2001 are merged into x $$ x $$ for 2011, v , w t 01 x t 11 $$ {\left\{v,w\right\}}_{(t01)}\leftrightarrow {x}_{(t11)} $$ . In this example, v $$ v $$ , w $$ w $$ and x $$ x $$ are connected and a consistent geography is produced by merging v $$ v $$ and w $$ w $$ : v w = x $$ v\cup w=x $$ . In fact, in these examples, all of v $$ v $$ , w $$ w $$ , x $$ x $$ , y $$ y $$ and z $$ z $$ are connected. The parent and cross-census neighbourhood for all of them is x $$ x $$ : v , w t 01 x t 11 y , z t 21 $$ {\left\{v,w\right\}}_{(t01)}\leftrightarrow {x}_{(t11)}\leftrightarrow {\left\{y,z\right\}}_{(t21)} $$ . More than 99% of cases are either unmodified across years or resolved by aggregation like this.

Complications arise when there is no single parent OA. Imagine OA v $$ v $$ is split into w $$ w $$ and x $$ x $$ , v t 01 w , x t 11 $$ {v}_{(t01)}\leftrightarrow {\left\{w,x\right\}}_{(t11)} $$ and that, subsequently, x $$ x $$ and another OA, y $$ y $$ , which was not a part of v $$ v $$ , are merged into z $$ z $$ , x , y t 11 z t 21 $$ {\left\{x,y\right\}}_{(t11)}\leftrightarrow {z}_{(t21)} $$ . Neither v $$ v $$ nor z $$ z $$ alone can be designated as the parent OA here. Instead, the cross-census neighbourhood is an amalgamation of v $$ v $$ , w $$ w $$ , x $$ x $$ , y $$ y $$ and z $$ z $$ , plus any other OAs that are connected to one or more of the set, at any of the three time periods, as well as any connections of those connections, and so forth, recursively, until no additional connections are found.

In a very small number of instances (57, about 0.03%), there is no reported link between an OA in 2011 and any originator in 2001. These were resolved through proximity matching; by connecting the population weighted centroid of the 2011 OA to the nearest population weighted centroid of an OA in 2001. The final outcome is a set of 174,363 cross-census neighbourhoods for England and Wales, of which 167,232 (96%) are identical to the 2021 Census OAs. Further information can be found at https://profrichharris.github.io/census/harmonised/.

Ways to link censuses through time are not new (Lloyd et al., 2017; Martin et al., 2002) and they do not resolve all concerns about the comparability of the data. Sabatar and Simpson (2009) argue that comparisons of ethnicity data from successive censuses to examine population change ‘are often impossible, wrong or misleading. Distortions become more severe as the scale of areal units becomes smaller’ (p. 1461). They were writing in reference to the 1991 and 2001 censuses, for which their very different output geographies, as well as their differing ethnic categorisations, added to the inconsistencies in ways that are less problematic for the time period, used here, when both the geography and the ethnic categories are largely consistent across censuses. Nevertheless, it is only the geography that is harmonised, with no changes to the underlying data. Those are taken at face value even though they are subject to different non-response rates and different imputation methods for trying to resolve undercounts, item non-response and item inconsistencies (see, for example, Martin, 2010).2 The same trust is made of the meaningfulness or otherwise of the ethnic categories, although they are not as stable as would be hoped (Simpson et al., 2016) and are clearly a somewhat crude simplification of the complexity of ethno-cultural identities.

3.2 Measuring diversity

A well established measure of ethnic diversity is the entropy index (Theril & Finizza, 1971). The formula for this, scaled so it ranges between zero and a theoretical maximum of one is,
E i = 1 / log m k = 1 m p ki log p ki $$ {E}_i=-1/\log (m){\sum}_{k=1}^m{p}_{ki}\log \left({p}_{ki}\right) $$ (1)
where i $$ i $$ denotes a neighbourhood in a city, k $$ k $$ is any one of the ethnic groups under consideration, m $$ m $$ is the number of ethnic groups and p ki = n ki / n + i $$ {p}_{ki}={n}_{ki}/{n}_{+i} $$ , where n ki $$ {n}_{ki} $$ is the number of ethnic group, k $$ k $$ , residing in i $$ i $$ , and n + i = k = 1 m n ki $$ {n}_{+i}={\sum}_{k=1}^m{n}_{ki} $$ , which is the total number of all m $$ m $$ groups in the neighbourhood.

The diversity score is calculated for each neighbourhood, in each of the four cities, at each of three time periods: 2001, 2011 and 2021. For these, m $$ m $$ (in Equation 1) is 14 and the groups are, in ascending order of their total number across the four cities in 2021: the White British group (37.9% of the total); the White Other group (excludes the White Irish) (12.8%); those of Indian heritage (8.0%); those of Black African heritage (7.7%); the ‘Other’ group (including Arabs, although these could be identified as a separate group from 2011 onwards) (6.0%); Pakistani (5.2%); Asian Other (4.2%); a ‘mixed’ or multiple ethnicity that includes identifying with being white (3.8%); Black Caribbean (3.7%); Bangladeshi (3.6%); any other mixed or multiple ethnicity (1.8%); White Irish (1.7%); Chinese (1.6%); and Black Other (1.6%). In 2001, the White Other group includes those of a gypsy/traveller/Roma community. From 2011 onwards they are counted separately but omitted from the analysis, from the White Other group, and from the total population count because they are a much smaller group than the rest (0.5% of the total) with, traditionally, different residential patterns of living. This means that the White Other group for 2001 as it is applied here is not strictly comparable with that of 2011 and 2021, but the presence/absence of the gypsy, traveller and Roma groups has negligible effect on the totals. The decision to collapse the various mixed ethnicity groups that identify as white into a single group was pragmatic: individually the various sub-groups (White and Asian, White and Black Africa, and White and Black Caribbean) form small percentages of the total populations; combining them reduces the amount of information to be displayed in the subsequent charts.

The diversity score is zero when the entire population of a neighbourhood belongs to a single group. It is one when all the groups are equally sized; that is, when n ki $$ {n}_{ki} $$ = n + i / m k $$ {n}_{+i}/m\kern0.5em \mathbf{\forall}\mathrm{k} $$ . As a measure of ‘complete diversity’ it is somewhat contrived both because some of the groups are rather amorphous (‘Asian Other’, for example) and because the groups' different demographic histories and levels of representation in the population mean that the possibility of them constituting one-fourteenth of a neighbourhood's population has varying degrees of plausibility. Nevertheless, it is widely used and although an alternative would be the Reciprocal Diversity Index (RDI) used by Catney et al. (2023), in practice they are similar measurements: the Pearson correlation between the E i $$ {E}_i $$ and RDI i $$ {RDI}_i $$ scores for the four cities, in 2021, is 0.92. The correlation between E i $$ {E}_i $$ and log RDI i $$ \log \left({RDI}_i\right) $$ is 0.98.


The diversity scores for the four cities at the three time periods are mapped in Figure 1, where it is evident that the cities have increasing and spreading diversity. In 2001, the diversity score of the average Birmingham resident's neighbourhood was 0.376, rising to 0.514 in 2011 and 0.588 in 2021. The corresponding values for Leicester are 0.346, 0.494 and 0.546; London, 0.498, 0.630 and 0.688; and Manchester, 0.353, 0.521 and 0.592. In Birmingham, 94% of its 3091 cross-census neighbourhoods have a higher diversity score in 2021 than in 2001. In Leicester, it is 85% of 885; London, 95% of 23,910; and Manchester, 93% of 1322. These cities and their neighbourhoods are becoming more ethnically diverse.

Details are in the caption following the image
The increasing ethnic diversity of neighbourhoods in Birmingham, Leicester, London and Manchester, 2001–21.

For most ethnic groups, this process of diversification means that each is residing, on average, in more diverse neighbourhoods in 2021 than in 2001, with the greater change happening between 2001 and 2011. This is shown in Figure 2. Among the four cities, London has the most diverse neighbourhoods, on average, and Leicester the least. In Leicester there is evidence of a slight decline in diversity of the neighbourhoods of the average Bangladeshi and Indian resident, but these remain more diverse than is typical for England as a whole, with the mean score for English neighbourhoods being 0.308, the median 0.227, and the interquartile range from 0.116 to 0.474.

Details are in the caption following the image
The neighbourhood diversity score for the average member of each of the ethnic groups in the four cities in 2001, 2011 and 2021. BAFR = Black African; BCAR = Black Caribbean; BOTH = Black Other; BANG = Bangladeshi; INDN = Indian; PKST = Pakistani; AOTH = Asian Other; CHNE = Chinese; WBRI = White British; WIRI = White Irish; WOTH = White Other; MIXW = ‘Mixed’ White; MIXO = ‘Mixed’ Other; OTHR = Other ethnicity.

Even the least diverse neighbourhoods in one census usually become more diverse in the next. Across all the neighbourhoods in the four cities, of the 7302 in the lowest quartile for diversity in 2001, only 4% had a fall in diversity for 2011. Of those in the lowest quartile for 2011, 5% had a fall in diversity for 2021. In total, just 437 of all 29,208 neighbourhoods (1.5%) had a fall in their diversity score from 2001 to 2011, and again from 2011 to 2021. Among those that did, 65 are in Brent, London (in places such as Alperton, where up to 66% of the population are of Indian heritage), 52 in Birmingham (including Sparkhill, up to 74% Pakistani), 47 in Leicester (including Belgrave, up to 90% Indian), and 42 in Newham, London (including Plashet, up to 35% Indian, and Manor Park, up to 40% Bangladeshi).

Table 1 examines the propensity for either the White British or another ethnic group to live in a neighbourhood where their group is the predominant or majority group. By predominant, it is meant that the group is living in a neighbourhood where their ethnic group comprises more than 80% of the population, indicating a strong concentration of that group in their neighbourhood. By 2021, neighbourhoods where a single group other than the White British formed more than half of the residents are rare in both London and Manchester, although their number has increased slightly: from 1.2% of neighbourhoods in London in both 2001 and 2011, to 1.7% in 2022; and from 0.9% in Manchester in 2001, to 1.5% then 4.5% in 2021 (see Table 1). In London, those few neighbourhoods are most likely to be majority Indian or majority Bangladeshi, but there are 40 times more White British majority neighbourhoods. In Manchester, they are most likely to be majority Pakistani, but there are 13 times more White British majority neighbourhoods. In 2021, there is just one neighbourhood among 23,910 in London that is more than 80% Indian and only two that are more than 80% Bangladeshi. There are 532 (2%) that are more than 80% White British. In Manchester, 103 of 1322 neighbourhoods (8%) are more than 80% White British. None have the same predominance of any other group.

TABLE 1. Percentage of neighbourhoods where the White British group and any other one of the 14 ethnic groups constitutes a majority of residents, and where the White British group and any other group constitutes more than 80%.
Form a majority Exceed 80% of the population
White British Another ethnic group White British Another ethnic group
Birmingham 77.2 6.3 51.4 0.2
Leicester 71.5 19.9 39.4 2.9
London 68.5 1.2 21.5 <0.1
Manchester 87.4 0.9 54.8 0
England & Wales 94.1 0.7 82.9 <0.1
Birmingham 65.9 (−11.3) 6.4 (+0.1) 27.9 (−23.5) <0.1 (−0.1)
Leicester 55.4 (−16.1) 21.8 (+1.9) 13.8 (−25.6) 4.0 (+1.1)
London 44.3 (−24.2) 1.2 (≈) 8.3 (−13.2) <0.1 (≈)
Manchester 74.0 (−13.4) 1.5 (+1.6) 24.4 (−32.4) 0 (=)
England & Wales 89.0 (−5.1) 0.8 (+0.1) 73.0 (−9.9) 0.1 (+0.1)
Birmingham 53.4 (−12.5) 8.7 (+2.3) 11.7 (−16.2) 0.1 (+0.1)
Leicester 39.0 (−16.4) 27.3 (+5.5) 1.5 (−12.3) 7.9 (+3.9)
London 30.7 (−13.6) 1.7 (+0.5) 2.2 (−6.1) 0.1 (+0.1)
Manchester 56.7 (−17.3) 4.5 (+3.0) 7.8 (−16.6) 0 (=)
England & Wales 84.8 (−4.2) 1.2 (+0.4) 63.2 (−9.8) 0.1 (≈)
  • a Values shown in brackets are the percentage point increase/decrease from 2001 to 2011.
  • b Values shown in brackets are the percentage point increase/decrease from 2011 to 2021.

In Birmingham, majority neighbourhoods of a group other than the White British are more common, but still rare: 6.3% of neighbourhoods in 2001, staying approximately equal in 2011, and rising to 8.7% in 2021. In 2021, 247 neighbourhoods of 3091 (8%) are majority Pakistani, with a further 9, 8 and 4 majority Black African, Indian and Bangladeshi, respectively. However, 1650 (more than half) are majority White British and 363 (12%) are more than 80% so. Only three neighbourhoods are more than 80% Pakistani and none is more than 80% Black African, Indian or Bangladeshi.

Among the four cities, majority neighbourhoods of a group other than White British are most common in Leicester, constituting 19.9%, 21.8% and 27.3% of the total 885 neighbourhoods in 2001, 2011 and 2021, respectively. In 2021, 241 neighbourhoods (27%) were majority Indian, one majority Black African and 345 (39%) majority White British. In 2021, Leicester is the only city with a greater number of neighbourhoods where more than 80% of their population is of an ethnicity other than White British than it has neighbourhoods that are more than 80% White British: 70 of the 885 neighbourhoods (8%) versus 13 (1%). That 70 has risen from 26 in 2001 and 35 in 2011.

Despite the general trend of diversification, residential enclaves of ‘minority’ groups do exist in all four cities and the number of those neighbourhoods has increased. However, they remain rare in comparison to where the White British live, except in Leicester where they are more common and may be reflective of members of the Indian and Bangladeshi groups living in neighbourhoods of slightly decreased diversity.

The general diversification of their populations could explain why the White British now form a lower percentage of the residents in each of the four cities. The change is quite marked, especially in Manchester where the percentage of neighbourhoods in which the White British predominated fell from 54.8% in 2001 to 7.8% in 2021. In 2001, the White British comprised more than 80% of the population in 1590 of Birmingham's neighbourhoods (51%), 349 of Leicester's (39%), 5140 of London's (21%) and 724 of Manchester's (55%). By 2021 the figures were 363 (a 77% decrease), 13 (96% decrease), 532 (90% decrease) and 103 (86% decrease), respectively. It is not just the percentages but also the numbers of the White British that have declined. Consequently, it may not be diversification that is creating smaller percentages of the White British but declining numbers of the White British that helps drive the diversification because, with the largest group becoming less dominant, the sizes of each group can become more equally balanced.

That possibility raises the question of where the rate of decline is greatest and whether any variation across neighbourhoods is related to how diverse they were initially—the basis for the ‘white avoidance’ hypothesis discussed earlier (although, as discussed in the Literature Review, it is, at best, circumstantial evidence due to other demographic factors). Figure 3 plots the ratio of the number of White British per neighbourhood in 2021 to their number in 2001, relative to how diverse the neighbourhood was in 2001. It is, in essence, a scatterplot but, in order to avoid over-plotting, it is shown as a hexagonal ‘heatmap’, where the shading of each hexagon indicates the number of data points within it. This choice of visualisation makes no difference to the weighted quantile regression lines that are also displayed, which estimate the conditional relationship at the first, second (median) and third quartiles, with a polynomial relationship and weights equal to the number of White British per neighbourhood in 2001. The gridlines on the x-axis show the interquartile and median diversity scores in 2001, pooled across all cities. The y-axis is on a log scale, slightly truncated because the full range is from 0 to infinity.

Details are in the caption following the image
Ratio of the number of White British (WBRI) residents in each of the cities' neighbourhoods in 2021 and 2001, relative to the diversity of those neighbourhoods in 2001.

It is evident from Figure 3 that more neighbourhoods in the four cities have a lower number of White British living in them in 2021 than in 2001 and that, generally, the most diverse neighbourhoods in 2001 lost a greater fraction of their White British population by 2021. However, there are many neighbourhoods, in each of the cities, that were diverse in 2001 and where the White British population subsequently increased, albeit sometimes from a low starting number.

4.1 Is segregation of the White British growing across the boundaries of the no majority cities?

Ethnic diversification of a population is not a guarantee against residential segregation. On the contrary, if everyone is of the same ethnicity then there is no possibility of different ethnic groups living in different places. The previous analyses established that all groups are living in more ethnically diverse neighbourhoods in the four cities, on average, but that does not preclude segregation occurring across the boundaries of the cities if, for example, the declining number of White British is associated with them moving from the urban cores to surrounding suburban and rural settlements.

If that is what is happening—and recall that there are actually demographic factors at play here too, not just processes of internal migration—then the places they are moving to are becoming more ethnically diverse too. This is shown by looking at the wider Travel to Work Areas (TTWAs) around each city, omitting the cities themselves. These TTWAs are based on 2011 Census data (Coombes & ONS, 2015), but remain useful to extend the spatial extent of the analysis to areas surrounding the cities that can then be grouped into different settlement types using a rural–urban classification, again for 2011 (DEFRA, 2016). Comparing Figure 4 with Figure 2 and noting the different ranges of the y-axes reveals that the settlements around the cities' boundaries are less ethnically diverse than the cities for where the average White British resident is living, especially in rural towns and villages. However, in all settlement types, around all four cities, the neighbourhood of the average White British resident has had increased ethnic diversity from 2001 to 2011 and again to 2021, with the greatest rate of change being around London, possibly because of affordability issues coupled with pressures of higher density housing in the capital that increase the propensity to leave if employment patterns permit it (Johnston et al., 2016a).

Details are in the caption following the image
The neighbourhood diversity score for the average member of the White British (WBRI) group living in a settlement in the TTWA surrounding the boundary of each city.
Figure 5 confirms that the White British are not becoming more segregated from other groups. It uses the classic Index of Dissimilarity (Duncan & Duncan, 1955) to calculate pairwise segregation scores between the named ethnic group and White British for each city and for each of the census years 2001, 2011 and 2021:
ID X ~ Y = 0.5 × i = 1 N n Xi n X + n Yi n Y + $$ {ID}_{X\sim Y}=0.5\times {\sum}_{i=1}^N\left|\frac{n_{Xi}}{n_{X+}}-\frac{n_{Yi}}{n_{Y+}}\right| $$ (2)
where X $$ X $$ and Y $$ Y $$ denote two different ethnic groups ( Y $$ Y $$ is always the White British in Figure 5), n Xi $$ {n}_{Xi} $$ is the number of ethnic group X $$ X $$ in neighbourhood, i $$ i $$ , n Yi $$ {n}_{Yi} $$ is the corresponding value for Y $$ Y $$ , and n X + $$ {n}_{X+} $$ and n Y + $$ {n}_{Y+} $$ are the total number of X $$ X $$ and Y $$ Y $$ across all N $$ N $$ neighbourhoods in the calculation. The index ranges from zero—where the share of the total of X $$ X $$ is the same as the share of the total of Y $$ Y $$ in all neighbourhoods, meaning they are equally distributed (‘no segregation’)—to one, which occurs if where X $$ X $$ is living, Y $$ Y $$ is not (‘complete segregation’).
Details are in the caption following the image
Index of Dissimilarity (ID) scores between the group (e.g., Bangladeshi, BANG) and the White British (WBRI) in 2001, 2011 and 2021 at each of five scales: (1) within the boundary of the city; (2) within the boundary of the Travel to Work Area of the city (TTWA0); (3) within the boundary of the TTWA and contiguous TTWAs (TTWA1); (4) within the boundary of TTWAs and their first and second-order neighbours (TTWA2); and (5) across the whole of England. BAFR = Black African; BCAR = Black Caribbean; BOTH = Black Other; BANG = Bangladeshi; INDN = Indian; PKST = Pakistani; AOTH = Asian Other; CHNE = Chinese; WBRI = White British; WIRI = White Irish; WOTH = White Other; MIXW = ‘Mixed’ White; MIXO = ‘Mixed’ Other; OTHR = Other ethnicity.

The Index of Dissimilarity is calculated for a range of spatial extents in Figure 5 to allow for the possibility that what is happening within the boundaries of the city is not reflected across wider geographical areas. The spatial extents are (1) the boundary of each city, (2) the city and their TTWA, (3) the city, their TTWA and neighbouring TTWAs, (4) the same as (3) but also including the neighbours of neighbouring TTWAs, and (5) the whole of England. For brevity, Figure 5 focuses on the largest of the Indian subcontinent, Black and White groups.

Looking at the results, as the geographical ‘net’ is cast wider from (1) through to (5) so the segregation value generally increases. This happens because a wider area includes more suburban and rural areas and these, as well as England as a whole, are not as ethnically diverse as the four cities. There are examples of the residential segregation remaining high in 2021. If 0.75 is taken as a threshold—above this, more than 75% of the second group would have to exchange places with the White British to achieve ‘no segregation’—then it is exceeded in a few cases and especially between the Bangladeshi and the White British at a range of scales in and around Birmingham, Leicester and Manchester, in particular. Nevertheless, in almost every instance, the segregation score has decreased between 2011 and 2021, the only exception being between the White British and Black African groups in the TTWA of Birmingham, but where the increase is trivial.


This paper has presented a study of England's only four cities with a population that is not majority White British and where no other ethnic group forms a majority. These cities are Birmingham, Leicester, London and Manchester. Of the cities, London has the highest average neighbourhood diversity and Leicester the least, but all four are much more diverse than England as a whole. Residential enclaves of ‘minority’ groups do exist within the four cities, wherein a single group forms the majority or more of a neighbourhood's population, and the number of such neighbourhoods has increased. However, they remain rare in comparison to where the White British live, except in Leicester where members of the Indian and Bangladeshi groups are living in neighbourhoods of slightly decreased diversity. Generally, neighbourhoods within these cities that were least diverse in 2001 have become more diverse. A tiny number had a decrease in diversity from 2001 to 2011, and again from 2011 to 2021. Overall, it is clear that the residential neighbourhoods of these cities have increased their ethnic diversity, with the average resident of each of 14 ethnic groups now living in a more diverse neighbourhood in 2021 than they did in 2001 or 2011.

When Nigel Farage3 (wrongly) Tweeted that ‘London, Manchester and Birmingham are now all minority white cities’ [London and Birmingham are not], the Conservative MP and former Chancellor and Home Secretary, Sajid Javid, replied ‘so what?’. It is a good question. Placing emphasis on how ‘white’ cities are and, by extension, upon where the White British are living, risks promoting the idea that this group should be of particular importance to society, and contributing to a racialised nomenclature that reifies the idea of a native white population group, defined by skin colour and nationality, while concealing its internal and genealogical diversity. An answer might be that the ‘loss’ of the White British from the cities contributes to rising ethnic residential segregation and perhaps also decreasing diversification beyond their borders. This study, however, concurs with Catney et al. (2023) and finds no evidence of either occurring, despite the notable decline in the number of the White British population within the four cities.


My thanks to the reviewers of this paper for their helpful and insightful comments.


  1. 1 https://geoportal.statistics.gov.uk.
  2. 2 See https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationestimates/methodologies/itemeditingandimputationprocessforcensus2021englandandwales for information about the item editing and imputation process for Census 2021, England and Wales, and https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationestimates/methodologies/measuresshowingthequalityofcensus2021estimates for measures showing the quality of Census 2021 estimates.
  3. 3 Leader of UKIP, 2006–9 and 2010–16, and of the Brexit Party / Reform UK, 2019–21.

    The geographically harmonised data used in the study are available at https://profrichharris.github.io/census/harmonised/.