Ethnic identity and change 1991-2001

Ludi Simpson, University of Manchester

[Project number 40006]

Overall research aims:

  • To establish the nature and extent of an individual's change in recorded ethnicity from one census to the next, by estimating a transition matrix from ethnic group in the 1991 census to ethnic group in the 2001 census.
  • To associate changes and stability of ethnic group with personal, household and social characteristics.

Objectives and methods:

  • To review the distributions of ethnicity derived from the census at 1991 and 2001: complete population estimate, census without imputation, LS cross-section, LS linked.
  • To estimate the transition matrix from 1991 ethnic group to 2001 ethnic group, from linked data. To measure its reliability and consider improvements to the matrix, the given differences in distributions between the linked data and more complete sets. This will immediately offer answers to questions including
    What proportion of people changed their ethnicity?
    Where there ethnic groups where change was more likely?
    How did people who classified themselves as mixed in 2001, classify themselves in 1991?
    Are the changes plausible?
    A similar transition matrix between sex in 1991 and sex in 2001, and another between Country of Birth in 1991 and in 2001 will provide a measure of the random noise that might be expected in LS data with repeated measures, due to errors in data recording and data capture.
  • Attempt to explain key parts of the transition matrix in relation to other variables and changes in those variables: age, cohort and period are all expected to influence the choice of census ethnic group, as well as educational, household and employment circumstances. For example:
    Age, sex, religion and household composition: are these associated with changes in ethnic group? Is a young child's assigned ethnic group likely to change by early teens, on reaching adult age, on leaving home? For religion, an analysis of the contingency table of ethnicity 1991 vs ethnicity 2001 vs religion 2001.
    Parents of different ethnicity: how were their children classified, in 2001 and in 1991?
    Household ethnicity: is a change in individual ethnicity associated with the current household composition?
    Ward ethnicity: are individual changes in ethnicity associated with the current ethnicity of the neighbourhood? Are they associated with migration between neighbourhoods of different ethnicity?
    NS-SEC, qualifications, economic activity: are changes in these individual variables associated with changes in ethnic group.
    Geographical differences: devising a measure of overall change in identity, what is the variation between them? We will pay specific attention to those ethnic groups which have been most unstable, and to new categories. These are likely to include the Irish, White Other, Mixed and the Black Caribbean groups.

The analysis will begin with exploratory summaries of the data. At each stage, the plausibility of the results will be used to question the data quality.

The complexity of the tables will also require modelling, using contingency tables, logistic and multinomial logistic regression. The depth of the analysis during this beta-test research will be limited by the time available.

Top of page