Critical responses to homogenization of ethnic analysis: the 1991 Census ethnic group categories

In Britain the most pervasive ethnicity categorisation of contemporary quantitative research is that of the 1991 census, developed through a considerable exercise in coding and categorising textual responses to questions on subjective ethnic identity (Bulmer 1996, Sillitoe & White 1992). The census schema’s attraction lies in its apparent ability to produce a parsimonious and theoretically coherent account of ethnic difference. Thus on the one hand it produces a demographic map showing a handful of large, discrete and putatively homogeneous ethnic groups, principally identified with historically different waves of immigration (Peach 1997). On the other, it makes reference to a “popular” discourse on the meaning of ethnicity, namely individuals’ ethnic identities, derived from the contextual fusion of elements such as identification with country of ancestors’ origins, skin colour, language, shared cultural values, shared participation in the social conditions of a group, and religion (cf Anthias & Yuval-Davis 1992, Banton 1997, Mason 1995). The resulting categories reflect a trend of “origins-privilege” in British conceptualisations of ethnicity. For instance, a common pattern in sociological discussions mimics the development of the census categorisations : a conception of subjective ethnic identification, which is in theory unconstrained, is eventually (and often begrudgingly, cf Anthias 1998), whittled down to a discussion in terms of categories heavily coincident with the history of immigration to Britain (Mason 1995, Modood et al 1997, Solomos 1993)3. Subsequently, the categories of the 1991 census have come to dominate the terms of contemporary social research, been replicated in both government and private surveys, and adopted in the monitoring systems of major institutions (Coleman & Salt 1996a). Table 1 below shows the ‘ethnic map’ of Britain produced by the 1991 census categories.


Table 1
The UK population categorised by selected ethnic groupings
(numbers or estimated numbers in millions)
1991 Census Religion 2001 Census
White (51.9) Christian / none / other (53.1) White-British
Black-Caribbean (0.50) Muslim (1.05) White-Irish
Black-African (0.21) Hindu (0.39) White-Other
Black-Other (0.18) Sikh (0.34) Mixed White/Black-Caribbean
Indian (0.84) Mixed White/Black-African
Pakistani (0.48) Indian Sub-Division Mixed White/Asian
Bangladeshi (0.16) Indian sub-continent (0.50) Mixed Other
Chinese (0.16) African-Asian (0.34) Black-Caribbean
Other Asian (0.20) Black-African
Other Other (0.29) Black-Other
Languages spoken
Indian
English / Celtic only
(52.6) Pakistani
Immigrant Status
Non-European
(2.3) Bangladeshi
Born in UK (51.1)
Other Asian Born abroad (3.7)
Chinese
Any other


Sources :
· 1991 Census and Immigrant Status : Coleman and Salt (1996b:134)
· Religion, Language and Indian Sub-Division : approximations derived from Modood et al (1997:298,309) (1994 PSI survey) compared against 1991 Census proportions4



However, a number of critical perspectives can be presented on the 1991 census
categorisations, which when combined leave limited theoretical support to justify their considerable research influence.

Initially, several positions are theoretically congruent with the census categorisations 5, but find inadequacies in their practical implementation. Firstly, pragmatic support for the style of operationalisation has been countered with the claim that the categories are badly chosen, and moreover will be increasingly inappropriate over time (eg Owen 1996) – an influential argument which resulted in modifications made to the schema for the 2001 Census (UK Command Papers 1999, and see table 1.1). Whilst some such positions advocate the mere adjustment of the appropriate category labels, a more fundamental critique anticipates that social, cultural and demographic trends in Britain mean that the census categories’ Weberian conflation of ‘origins’ with ‘identities’ is rapidly loosing purchase over time (Berthoud 1998a).

A second, related, critical position holds that the 1991 census categories give an inappropriate weight to multiple alternative (but measurable) ethnic referents, such as those shown in table 1.1. Whilst the 1991 census categories broadly coincide with many of these boundaries, it may be plausible to claim that specific applications require a stronger focus upon alternative referents6, or perhaps their combination with subjective ethnic identity categories. Certainly, the incorporation of categories cross-classifying referents such as subjective identity, language, religion and country of birth may not be as complex for Britain as may be expected, since there is a considerable degree of clustering into categories.

One important point, however, concerns the subsequent confusion caused if different research traditions favour the use of alternative or combined ethnicity referents. This is aptly illustrated in attempts at cross-nationally comparative discussions of ethnicity effects, where the common solution of making do with which ever referents are nationally available (Lambert & Penn 2001, Stille 1999), causes confusion and cynicism over the strength of comparative conclusions (Hoffmeyer-Zlotnik 2000, Lloyd 1995).


1 This paper is constructed primarily from results presented in Lambert (2002), which discusses many of the findings presented here in greater detail.

2 We concentrate here upon the example of Britain, but even a cursory review of the social science research situations of other countries reveals similar confrontations are widespread (cf Alund & Granqvist 1995, with Stille 1999).

3 In some cross-nationally comparative research a similar neo-Weberian conception of ethnicity is advocated, as in the concept of country specific “ethno-national identities” used in the works in Martiniello (1995). Panayi’s (1999) study reflects how, as in Britain, most such derived categories ultimately reflect patterns of ancestral origins.



4 The information on ‘language spoken’ estimates those with the ability to speak a non-European language, not those who necessarily speak such a language regularly. The study by Modood et al (1997) does not cover all relevant census categories, prompting working approximations based upon the descriptive evidence of Peach (1997). Thus we treated the Black-Other group with the same proportions as the Black-Caribbean groups. It is also estimated that 50% of the Black-African, Other-
Asian and Other-Other populations followed ‘Christian / none / other’ religions, and the other 50% were Muslim; and that 90% of the members of all of those groups could speak a non-European language.

5 In their support of the principal that subjective ethnic identity could adequately summarise the constellation of ethnic elements mentioned above, these positions could be described as Weberian, cf Banton (1997).

6 For instance, although Modood is generally a proponent of the maximum number of subjective ethnic categorisations (Modood et al 1994, Modood et al 1997), in one significant article he suggests distinguishing Asian groups solely between Muslim and non-Muslim for the analysis of economic attainment (Modood 1991).

Derived from:
Lambert, P., 2002,“Quantitative Representations of Ethnic Difference?”
The University of Oxford;
http://www.nuff.ox.ac.uk/rc28/Papers/lambert.pdf