Although most surveys contain an ethnicity question, many national surveys contain only small numbers of people from minority ethnic backgrounds.
This often means that data for some ethnic groups cannot be analysed or published separately, due to statistical unreliability or disclosure control issues.
One way around this is to aggregate data - either by combining more than one year's data, or by combining different ethnic groups.
Assess which categories can be combined.
Ideally categories should only be combined if they show similar patterns on the outcome variable of interest.
You should not automatically combine some large and important categories just because one category is particularly small, as it isn't always appropriate.
For example, for female unemployment rates from the Annual Local Labour Force Survey, Bangladeshi information may be too small to be shown separately.
However, it is not appropriate to combine this data with 'Indian', 'Pakistani' and 'Other Asian', into one 'Asian or Asian British' figure, as the employment pattern for Indians is very different to that for Pakistanis and Bangladeshis. Instead, data for Indians could be shown separately and data for Pakistanis and Bangladeshis combined. Data for the 'Other Asian' group could be either omitted or added into the 'Other ethnic group' category.
Instead of combining categories inappropriately, it may be better to show a category in a table while indicating that data for that category have been omitted because of small sample sizes.
For example, for male unemployment rates from the Annual Local Labour Force Survey, sample sizes are big enough to show rates for 'Black Caribbean' and 'Black African' men separately, but too small to show the 'Other Black' group.
If you combine all three categories into one 'Black' group you will lose the distinction between 'Black Caribbeans' and 'Black Africans'. However, this can be avoided if each category is displayed separately but the data for the 'Other Black' men are not shown.
Homepage
