Scott OatleySociology / UCL
← Guides

Ethnicity in ELSA: an interactive guide

How to build consistent ethnicity measures across twelve waves of the English Longitudinal Study of Ageing

Based on Oatley, S. (2025) ELSA and Ethnicity: Detailed Roadmap of Ethnicity Data Collection in the English Longitudinal Study of Ageing (ELSA).

Ethnic inequalities in later life are under-studied, mainly because most surveys have too few older ethnic minority respondents to analyse. ELSA asks about ethnicity at every wave, but the questions, variable names and level of detail change from wave to wave. The public data hold only a White / Non-White split, and early values were fed forward from the Health Survey for England. The report builds two consistent measures, a dichotomised one and a detailed one, and tests where each has enough cases to use.

Detailed ethnicity is available to the ELSA team; the deposited UK Data Service files contain only the White / Non-White variable. Wave 11 figures use internal interim data.

1What each wave records

Click a cell to see the variable, its categories and how many people answered.

HSE fed from the Health Survey for England · ELSA collected by ELSA · detailed detailed categories

2Carrying values forward and back

About 90% of ELSA members take part in more than one wave (only 10.57% appear just once). If ethnicity does not change, a value recorded at one wave can fill gaps at others:

The dichotomised measure uses LOCF; the detailed measure uses both. The detailed measure then borrows “White” from the dichotomised measure, the one category the two share. Step through an example person.

One person’s detailed ethnicity across waves

3What it adds

Across all waves there are 132,294 person-wave observations. These are the counts with a usable ethnicity value at each step.

4Is ethnicity fixed?

Carrying answers across waves assumes a person’s ethnicity does not change. The report checks this by counting “identity switchers”: people whose recorded ethnicity differs between waves.

MeasureObservationsSwitcher observationsIndividuals
Dichotomised132,12856674
Detailed129,37326843

Of the 268 detailed switcher observations, 108 move between White and an ethnic minority group and 160 move between minority groups. Against more than 129,000 observations, switching is rare: ELSA members very largely report their ethnicity consistently, which supports the assumption.

5Ethnic minority share by wave

The share of Non-White respondents (dichotomised measure, after carrying values across waves) rises from wave 8 and reaches its highest level at wave 11. That follows ELSA’s efforts to boost ethnic minority numbers, including the wave 10 refreshment sample aged 50–93 and a revised ethnicity question matching the census.

Wave 0 is HSE-based data for a much larger sample (26,787).

6Which waves to use

The report sets a working threshold: a detailed measure is usable once every ethnic minority group exceeds 0.25% of valid responses in a wave, which happens from wave 4. Describe your analysis to get the report’s advice.

Design
Detail needed

7Check yourself

8The report