Content uploaded by Ruth Lupton
Author content
All content in this area was uploaded by Ruth Lupton
Content may be subject to copyright.
UNDERSTANDING POPULATION TRENDS AND PROCESSES
UPTAP
RESEARCH FINDINGS July 2010
present selected findings of analysis of 2008 Annual
School Census (ASC) data for the capital to quantify and
map the languages spoken in contemporary London,
updating Baker and Eversley’s work and showing how
the situation has changed. We show how an analysis of
language rather than ethnicity alone can shed new light
on patterns of educational inequality.
Key Findings
Although around 60% of London pupils are English
speakers, there are over 40 languages spoken by
more than 1,000 pupils. Bengali, Urdu and Somali are
the top three languages spoken in London, other than
the English language.
English has a ‘doughnut’ shaped geographical
distribution in London, being the predominant
language in most of Outer London. Languages other
than English are more common in Inner London. Most
minority languages, such as Bengali, Urdu and Turkish,
have one, two or three main clusters, reflected settled
immigrant communities. However others, notably
Somali, are widely dispersed. This has implications for
service provision.
Some of the ethnic categories that are widely used in
analysis of census data hide substantial linguistic
diversity, particularly ‘Black African’ and ‘White Other’.
Within London, where these groups are numerous,
language data provides a valuable disaggregation of
these heterogeneous groups.
Language spoken provides a means to better
understand the relationship between ethnicity and
educational performance.
The increasing linguistic diversity of the UK attracts
much interest and debate among public service
providers, educationalists and the public. The presence
of languages other than English has been seen as both
an asset and a liability in education (Mehmedbegovi
2007), an economic opportunity (Martin, 2000) and a
major economic cost (BBC, 2006), an expression of
multiculturalism and a threat to community cohesion.
More immediate pragmatic concerns include
establishing whether we have the language skills to do
business with the rest of the world, and where these
skills are located, where and in what contexts is there a
need for translation skills or language classes, and to
what extent the languages of the five London Boroughs
hosting the 2012 Olympics match those of our world
visitors.
Yet remarkably, little is known about the numbers of
people who speak different languages, and the
implications of this dimension of population
composition and change. Language data have not been
collected in the Census of Population, and until 2007,
School Census data only contained information about
English as an Additional Language (EAL) not the
individual languages. A ‘model’ question asking about
the language spoken at home was inserted in 2007
(DfES, 2006a; 2006b) and from 2008, all schools were
required to collect these data. Although this source only
provides information about school pupils, not about all
residents, it represents a major advance on our
knowledge thus far.
London is by far the most linguistically diverse part of the
UK and was the subject of an earlier analysis by Baker
and Eversley (2000), based on pupil data collected from
individual London boroughs, which identified over 300
languages spoken by London school children. Here, we
LANGUAGES, ETHNICITY AND EDUCATION IN LONDON
Michelle von Ahn London Borough of Newham
Ruth Lupton London School of Economics and Political Science
Charley Greenwood and Dick Wiggins
The Institute of Education, University of London
2
UPTAP RESEARCH FINDINGS
Data, language classification and geography
Our analysis draws on data from the Annual School Census in
2008. This covers pupils in all state schools, but not private
schools. Around 1.1 million pupils are represented in the
dataset. The data collection instrument provides a list of 322
languages, of which some are variants of other languages.
For example, a person may be classified as speaking
Bengali (main category) or Bengali (Sylheti) or Bengali
(Chittagong/Noakhali) or Bengali (any other). This structure
provides some inconsistencies as different local authorities
have collected data at different levels — some using main
categories only and some using variants — and for this
reason, analysis of the variants is not reliable. Furthermore,
there are even less detailed categories such as ‘believed to
be English’ and ‘other than English’. For London as a whole,
7% of records have this insufficiency of language detail and
there are tiny percentages of refusals and missing values
(in total 1.4%). However, at borough level, the ambiguity
can range from 2.3% (Ealing) to 27.9% (Westminster).
Clearly data quality issues need to be addressed if the data
are to enable sophisticated analysis of change over time.
We classify languages based primarily on their location in
the world (Dalby, 1999). Eight ‘geozones’ are identified, as
follows: Asia (South), Asia (East), Asia (West/central),
Africa (North), Africa (West), Africa (East/Central/South),
European Union and Other Europe. In addition, the
classification includes a category
‘international/transnational’, incorporating the major
languages of Arabic, French, Portuguese and Spanish,
which are spoken in many parts of the world, as well as
‘other’ languages. Figure 1 shows this classification,
highlighting languages which are spoken by more than
5,000 pupils in London’s state schools.
We use postcode data for individual pupils in the Annual
School Census to assign all records to Super Output Areas
(SOAs) (Vickers and Rees, 2007), which themselves exist in
varying levels of aggregation, i.e. ‘lower’ with around
1,500 residents and ‘middle’ consisting of around 7,500
residents known respectively as LSOAs and MSOAs. This
enables us to map languages by the home address of
pupils, whereas Baker and Eversley were previously only
able to calculate figures at borough level based on the
location of the school attended. For the London-wide maps
that follow, we show only data for London’s 983 MSOAs.
We show numbers of pupils rather than percentages, given
the unevenness of the data quality (denominators) across
the boroughs. For each language, five categories are
shown. It is important to note, therefore that the scales for
the maps are different to one another. The primary purpose
is to show the geographical distribution of each language,
not to compare volumes.
Patterns or clustering of language
English speakers, including a small percentage classified as
‘believed to be English’, (<1%), represent over 60% of
state pupils in London (663,584 in total). No other
language is spoken by more than 5% of all pupils with a
recorded language, and over 40 languages are spoken by
more than 1,000 pupils — a picture of remarkable diversity.
The 15 most prevalent languages other than English are
shown in Table 1. Column 1 shows the ordering of
languages in 1999 from Baker and Eversley (2000). The
table indicates that while overall there has been little
change in the relative importance of languages spoken in
the nine years since 1999, there are some differences.
Notably, Somali speakers have become more prevalent
along with Tamil speakers which may well represent recent
turmoil in their countries of origin. Polish and Albanian
speakers appear in the 2008 rankings for the first time in
comparison to Greek, Cantonese and Creole speakers,
possibly reflecting recent expansion in EU membership.
Maps of individual languages demonstrate patterns of
settlement and dispersal of minority communities and
provide a basis for understanding how these patterns
change between Censuses of Population. Figure 2 clearly
shows a doughnut pattern, with English speaking pupils
found more in outer London (particularly in the East and
Other
Asian
West/Ce ntral
Asian
North
Asian
West
Asian
E/C/S
European
Union
European
Other
International/
Transn ational
Asian
East
Asian
South Other Unspecified
Bengali Chinese Tu rki sh Somali Yor ub a Lingala Greek Albanian Arabic Caribbean
Creoles
Urdu Vietnamese Persian/
Farsi Tigrinya Akan/
Twi-Fa nte
Swahili /
Kiswahili Italian Russian French Oceania/S/
C/America
Panjabi Japanese Kurdish Amharic Igbo Luganda Dutch/
Flemish
Serbian/
Croatian/
Bosnian Portuguese
Gujarati Korean Pashto/
Pakhto Other Other Shona German Other Spanish
Tami l Tag alo g/
Filipino Other Other Bengali
Polish
Hindi Other Lithuanian
Malayalam
Nepali
Other
>5000
pupils
FIGURE 1. A CLASSIFICATION OF LANGUAGES
July 2010
3
South) than inner London. Minority language speakers tend
to be concentrated in particular parts of the city. For
example, Bengali speakers are heavily concentrated in
Tower Hamlets, and Urdu speakers in three main areas: the
neighbouring boroughs of Newham, Redbridge and
Waltham Forest, Ealing/Hounslow and Merton/Wandsworth
(Figure 3). However, note the much more dispersed
distribution of Somali speakers (a similar size population
overall to Urdu speakers) (Figure 4). Similar language maps
can be found in the new edition of Multlingual Capital
(Tinsley et al., 2010).
Language and ethnicity
While many languages ‘attach’ to particular ethnic groups,
there are others (those that we have classified as
‘international’) for which knowing a person’s language
does not tell us about their country of origin or ethnic
heritage. Data on ethnicity, using the 16 major ethnic
categories used in the Census of Population, is also
collected in the Annual School Census. This reveals that
57% of French speaking pupils are ‘Black’ and a similar
percentage of Arabic speakers are classified as ‘Other Black’
(15%), Mixed (10%), White (9%) or Asian (8%). This
suggests the need to analyse language and ethnicity data
in these cases to understand the nuances of people’s
circumstances and needs. Notably these different
populations have different geographical concentrations.
White French speakers tend to reside in West London, Black
French speakers in East London.
Correspondingly, language data can potentially offer a
finer-grained understanding than has to date been
available through the collection of ethnic categories. Some
ethnic groups are characterised by considerable linguistic
homogeneity. For example, 84% of pupils identified as
Bangladeshi in London speak Bengali at home (with a
further 12% categorized loosely as ‘other than English’, of
which some will be Bengali speakers). 98% of White
British and 95% of Black Caribbean children speak English
at home However, other ethnic groups are very
linguistically diverse, most notably ‘Black African’ and
‘White Other’. Around 30% of Black Africans speak English
Rank Year
1999
Year
2008
Number
2008
1. Bengali and Sylheti Bengali 46,681
2. Panjabi Urdu 29,354
3. Gujarati Somali 27,126
4. Hindu/Urdu Panjabi 20,998
5. Turkish Gujarati 19,572
6. Arabic Arabic 19,378
7. English based Creoles Turkish 16,778
8. Yoruba Tamil 16,386
9. Somali Yoruba 13,961
10. Cantonese French 13,020
11. Greek Portuguese 11,915
12. Akan Polish 10,991
13. Portuguese Spanish 8,647
14 French Albanian/Shqip 8,380
15. Spanish Akan 8,117
TABLE 1. THE ‘TOP 15’ LANGUAGES SPOKEN OTHER THAN ENGLISH
IN LONDON
FIGURE 2. DISTRIBUTION OF ENGLISH-SPEAKING PEOPLE IN LONDON
FIGURE 3. DISTRIBUTION OF URDU-SPEAKING PEOPLE IN LONDON
FIGURE 4. DISTRIBUTION OF SOMALI-SPEAKING PEOPLE IN LONDON
at home, 20% Somali, 9% Yoruba, 6% Akan, 5% French,
2% Lingala, 2% Igbo and 2% Arabic. There are 179 other
languages spoken by fewer than 2% each of the London’s
Black African pupils.
The main African languages spoken in London originate in
different parts of the continent. Yoruba, Igbo and Akan are
spoken mainly in West Africa, including Nigeria and Ghana.
Lingala is spoken in Central Africa. Among the ‘Other
White’ ethnic group, Turkish (14%) is the most common
language, but 10% speak Polish, 8% Albanian or Shqip, 6%
Portuguese, and 3% each Lithuanian, Greek and Spanish.
‘Indian’ is also a linguistically diverse category, with two
major groups in Gujerati (29%) and Panjabi (22.6%), as well
as Hindi, Urdu, Tamil and Malayalam speakers. For these
heterogeneous groups, the collection of data on language
provides an opportunity for finer grained understanding of
who is living in London and their socio-economic
circumstances, and how these are changing over time.
Language and attainment
The usefulness of ethnic/language categories is
demonstrated by a preliminary analysis of educational
attainment data. Here, for simplicity, we concentrate only
on results at Key Stage 2 (age 11). Using a 16 category
classification (DMAG, 2003; 2005) for ethnicity we see
considerable differences between ethnic groups (Figure 6).
Pupils of Chinese ethnicity are on average the highest
attainers, with a median of 15.38 points. Black Caribbean,
Black Other and Black African pupils are the lowest
attainers, with medians of 13.55, 13.67 and 13.73
respectively. Groups that are predominantly English-
speaking appear throughout this distribution, from Black
Caribbean at the bottom to White British and White Irish
near the top.
Figure 5 highlights the wide distribution of scores within
each ethnic group. The solid boxes in the chart show the
25th and 75th percentiles for each group, while the
‘whiskers’ show the range of attainment beyond this. Most
groups have a gap of about 2.5 points between the 25th
and 75th percentiles (slightly wider for the ‘White Other’
and ‘Other’ categories. However there are very high
attaining pupils in all groups.
Figure 6 and Table 2 illuminate the spread of attainment
within the two ethnic categories which had the greatest
linguistic diversity — White Other and Black African. Within
White Other, five groups have particular low attainment.
Median scores for Turkish, Portuguese, Lithuanian and Polish
speakers (as well as people whose language is classified as
‘Other than English’) would put them at the bottom of the
distribution in Figure 5. While there are high attainers in
these groups, there are also long tails of low achievement.
By contrast, Italian, Greek and English speakers in the
White Other ethnic category have few low attainers and
median scores that place them close to the top of the
overall distribution. Attainment patterns for White Greek
speakers are similar to those of people who identify as
having Indians ethnicity (a median of around 14.6 points).
The Black African category also contains a wide spread.
Table 2 shows the three lowest attaining and three
UPTAP RESEARCH FINDINGS
highest attaining Black African language groups, by
comparison with some of the main ethnic groupings. Note
that Lingala, French and Somali speakers tend to have
very low attainment, well below that of the lowest
attaining ethnic group overall (Black Caribbean). The
attainment of Black African Igbo speakers is similar to that
of White British students. These data suggest that some of
the commonly used ethnic groupings may be too broad to
be useful, and that language data can provide greater
insight into which pupils may be in need of particular
support.
Language, ethnicity and
socio-economic circumstances
These data give no indication that language itself is
responsible for greater or lesser attainment. It is necessary
to consider the different socio-economic circumstances and
migration histories of people who have come to London at
different times and from different parts of the world. The
4
Key stage 2 total (fine)
17.5
15.0
12.5
10.0
7.5
Black Caribbean
Black Other
Black African
Other
Pakistani
Bangladeshi
Mixed White & Black Caribbean
Other White
Mixed White & Black African
Any Other Mixed
White British
Other Asian
Inidian
White Irish
Mixed White & Asian
Chinese
FIGURE 5. KS2 TOTAL SCORES BY ETHNIC CATEGORY, LONDON 2008
Key stage 2 total (fine)
17.5
15.0
12.5
10.0
7.5
Turkish
Other than English
Portuguese
Lithuanian
Polish
Albanian/Shquip
Spanish
Greek
English
Italian
FIGURE 6. KS2 TOTAL SCORES BY LANGUAGE CATEGORY
(10 LARGEST), WHITE OTHER, LONDON, 2008
ASC data only contains limited fields of data on socio-
economic circumstances namely, whether or not a pupil
receives free school meals (FSM) and an index of
deprivation describing the pupil’s residential
neighbourhood (IDACI). To enrich our understanding of the
circumstances of different ethno-linguistic groups, we are
able to match and merge the ASC data on data for the
London Borough of Newham with data for that borough
previously matched by our consultants, Mayhew Harper
Associates Ltd (2009).
At a local authority level, data records are linked together
via a property gazetteer using the General Practice registers
as a base for reference. Addresses are cross-referenced and
checked as to whom is present using various logical
assumptions to include or exclude people.
Analysis of these data shows that, for Newham at least,
there are marked differences in socio-economic
circumstances within language and ethnic groups, which
may well be driving attainment patterns (Table 3). On both
poverty indicators FSM and whether the family is in receipt
of Council Tax Benefit (CTB) Black African Somalis are by far
the most disadvantaged group. They also have the highest
proportion of single parent families and larger families.
Other ethno-linguistic groups with large families tend to
have low proportions of single parents, and vice versa.
Ethnicity alone certainly gives a misleading picture. Yoruba
speakers are relatively advantaged on these measures, in
great contrast to Somalis. On the other hand, considering
language alone would also be inadequate. White British
English speakers in Newham appear more disadvantaged
than Black Caribbean English speakers, and Pakistani
Panjabi speakers more disadvantaged than Indian Panjabi
speakers.
Next steps
These findings highlight the potential of the ASC language
data to help disaggregate Census ethnic categories and
give greater insight into the geographic distribution and
socio-economic circumstances of different ethno-linguistic
communities. Annual analysis of the data could provide a
vital inter-censal picture of settlement and migration,
providing that data is consistently and accurately collected.
The ASC remains a state school exercise. In some parts of
England and Wales and specifically in London, this is a
significant gap. For example, in Kensington and Chelsea
less than 50% of children are believed to attend local state
secondary schools. Some may attend state schools in
neighbouring areas. This is a reason for seeking to do
regional rather than local studies. The existence of
specialist private schools for speakers of other languages
such as the Lycee Francaise or the German School in South
West London may lead to specific gaps in the data but in
July 2010
5
TABLE 2. LOWEST AND HIGHEST ATTAINING LINGUISTIC GROUPS
WITHIN THE BLACK AFRICAN CATEGORY CONTRASTED WITH
SELECTED ETHNIC GROUP SCORES, LONDON 2008
Language/
Ethnic Group
(by median score)
25th
percentile median 75th
percentile
Black African — Lingala 10.56 12.58 13.78
Black African — French 11.18 13.01 14.27
Black African — Somali 11.40 13.02 14.31
Black Caribbean
(average) 12.23 13.55 14.62
Black African (average) 12.25 13.73 14.85
Black African — English 12.84 14.13 15.21
Black African — Yoruba 13.02 14.19 15.13
Black African Igbo 13.04 14.36 15.48
White British (average) 13.03 14.38 15.52
TABLE 3. SOCIO-ECONOMIC CHARACTERISTICS OF LARGEST ETHNO-LINGUISTIC GROUPS IN NEWHAM (KS2 ONLY)
Pupils 3 or more
children %
Single
parent %
FSM
%
Council Tax
Benefit %
Black African English/Believed to be English 85 58 21 32 38
Black African Somali 88 89 41 91 97
Black African Yoruba 65 72 15 25 26
Black African Akan 47 68 19 30 36
Black African Other than English/unknown 118 80 32 59 63
Pakistani Urdu 203 77 11 27 44
Pakistani Panjabi 53 85 15 30 53
Bangladeshi Bengali 428 83 7 36 64
White British English 366 50 39 45 61
Indian Gujarati 122 56 4 15 34
Indian Panjabi 52 56 8 23 31
Black Caribbean English 175 45 39 27 44
Other Asian Tamil 54 61 11 9 51
Other Asian Tagalog Filiipino 40 50 18 5 10
general the high percentage of children who do attend
state schools makes the ASC an invaluable source of data.
The richer insight that can be gained by matching the
ASC data to other local administrative data is clearly
shown by Table 3. In practice, negotiating access to the
data presents a major challenge both ethically and
technically as well as the need to ensure data security
and confidentiality. At least three agencies or providers
are involved, the DCFS, the Primary Care Trust and the
local authority. As a result a fundamental component of
our project has been to test the viability and value of
such an exercise in the context of a single London
borough namely, Newham.
Finally, in this document we have concentrated on
description only. A key question is whether the
attainment patterns of different linguistic groups can be
entirely explained by their socio-economic position, or
whether language (in itself or as a marker of previous
circumstances and experiences) has explanatory power in
attainment. We will be exploring this further using more
advanced statistical techniques and both KS2 and KS4
data.
Acknowledgements
We acknowledge contributions from John Brown, John
Eversley, Antony Sanderson, Les Mayhew and Dina
Mehmedbegovi.
References
Baker, P. and Eversley, J. (eds.) (2000) Multilingual
Capital: The Languages of London’s Schoolchildren and
Their Relevance to Economic, Social and Educational
Policies, Battlebridge Publications, London.
BBC (2006) Translation costing public £100m, BBC News,
12 December. http://news.bbc.co.uk/2/hi/uk_news/
6174303.stm (accessed 25 June 2007)
Dalby, D. (1999) The Linguasphere Register of the World’s
Languages and Speech Communities, The Linguasphere
Press, Hebron (Wales).
Department for Education and Skills (2006a) School
Census for Maintained Secondary Schools, CTCs and
Academies in England Preparation and Guidance for
2007, DES, London
Department for Education and Skills (2006b) Pupil
Language Data Guidance for Local Authorities on
Schools’ Collection and Recording of Data on Pupils’
Languages (in compliance with the Data Protection
Act), DES, London.
Data Management and Analysis Group (2003/9) 2001
Census Key Statistics: Ethnicity, religion and country of
birth, DMAG Briefing, Greater London Authority, London.
Data Management and Analysis Group (2005/39)
UPTAP
For a full list of UPTAP Research Findings, visit www.uptap.net
6
London’s changing population: diversity of a world city
in the 21st century, DMAG Briefing, Greater London
Authority, London.
Mayhew Harper Associates Ltd. (2009) Making use of
local administrative data for population estimates and
service planning, Gordon Square Quantitative Methods
Seminar, The Institute of Education, 27
March,(www.nkm.org.uk accessed 19 January 2010).
Martin, C. (2000) Foreword, in Baker, P. and Eversley, J.
(eds.) Multilingual Capital: The Languages of London’s
Schoolchildren and Their Relevance to Economic, Social
and Educational Policies, Battlebridge Publications,
London.
Mehmedbegovi D (2007) ‘Miss, who needs the
languages of immigrants?’: London’s multilingual
schools, in Brighouse, T. and Fullick, L. (eds.) Education
in a Global City: Essays from London, Institute of
Education, London.
Tinsley, T., Everseley, J. Mehmedbegovi, D. and Wiggins,
R.D. (eds.) (2010) Multilingual Capital, 2nd Edition, The
National Centre for Languages (CILT) and Institute of
Education, London (in press).
Vickers, D. and Rees, P. (2007) Creating the UK National
Statistics 2001 output area classification, Journal of the
of Royal Statistical Society (A), 170(2): 379-403.
Contact details of the authors
Michelle von Ahn
London Borough of Newham, 4th floor,
West Wing, Newham Dockside, 1000 Dockside Road,
London E16 2QU.
Email: michelle.von.ahn@newham.gov.uk
Ruth Lupton
Centre for Analysis of Social Exclusion,
London School of Economics and Political Science.
Email: r.lupton@lse.ac.uk
Charley Greenwood
London Centre for Leadership in Learning,
Institute of Education, University of London,
20 Bedford Way, London, WC1H 0AL.
Email: C.Greenwood@ioe.ac.uk
Dick Wiggins
Department of Quantitative Social Science,
Institute of Education, University of London,
20 Bedford Way, London, WC1H 0AL.
Email: R.Wiggins@ioe.ac.uk
