Menu Close
white hands reaching up to speech bubbles
Graphic farm / shutterstock

BAME groups are underrepresented in polls of public opinion – here’s why it matters

The question of systemic racism in Britain has been impossible to avoid over the past year. But what about the more basic question of ensuring that ethnic minority voices are heard?

People from Black, Asian, and minority ethnic (BAME) backgrounds make up 14% of the UK. However, BAME participants are often hugely underrepresented in “nationally representative” polls, with BAME representation ranging below 10%, and sometimes even 5% or lower.

Even more common is the problem of non-reporting. Many UK polls that sample the general population also appear to exclude any mention of ethnicity. This is the case whether the topic is health (recent polls on coping with lockdown, vaccination priorities, or NHS staff wellbeing), politics (government approval ratings, the budget and voting intentions), and even attitudes towards immigration or race (including whether it’s acceptable to make jokes about race). In these cases, it is not clear whether ethnic representation was adequate because there is no indication that ethnicity was measured at all.

Such polls are often reported widely in the media, and so provide important social cues for how we understand public opinion. In one recent example among many, The Guardian, The Express, and The Independent all reported a large drop in public approval for Boris Johnson and the Conservative Party, as well as figures such as “43% of the public want to postpone the lifting of COVID-19 restrictions”, based on a report that had no mention of ethnic representation.

An analysis including ethnicity might have revealed even higher support among some ethnic minorities for postponing lifting of restrictions, and differences in public approval. Or perhaps not – but the twin problems of underrepresentation and underreporting illustrates the way BAME perspectives and opinions can be marginalised, with far-reaching consequences for how we understand the “voice” of the nation.

Nationally unrepresentative?

A basic methodological requirement for an opinion poll is a “nationally representative” sample – that is, the people surveyed fairly reflect the national population.

Typically, this means a survey would be expected to match the population that participants are drawn from in terms of age, gender, socio-economic status, voting preferences and geographic spread. In other words, a survey wouldn’t be considered robust if it had half as many women as men, excluded certain age groups, counted only Conservative voters or canvassed only wealthy professionals. But many surveys of public opinion often do not even report ethnicity, and those that do often have far fewer ethnic minority respondents than they should.

Children on a school trip
About a third of British school children are BAME. Altosvic / shutterstock

To account for low numbers, data is sometimes weighted for ethnicity. This makes sense when used to adjust for minor discrepancies, but is an inadequate approach for boosting a sample that is simply too small. Correcting larger disparities through weighting (for instance, pretending the answers of two respondents represent those of 20 respondents) adds error, and risks producing inaccurate results.

There are a few reasons why ethnic minorities may be more difficult to reach for polling. BAME groups are on average younger, less likely to own their home, and therefore potentially more transient and difficult to reach by post or telephone than white British groups. In recent years, response rates for telephone polls have declined steeply to less than 10% in inner city areas, which often have more BAME people. Other factors such as language barriers and mistrust may also have a role.

However, these challenges are not insurmountable. Official census exercises have samples that reflect the ethnicity of the population, as do large longitudinal nationwide surveys such as the British Social Attitudes survey, some government research papers on public opinion, and dedicated surveys that focus explicitly on ethnic minorities.

Truly representative samples can be obtained when there is an imperative to do so, but this does not appear to be the default. The consequence is that BAME populations are underrepresented in polls which may then be used to inform decisions. The resulting mirror that we hold up to society is one that is distorted, and potentially blind to the opinions of key sections of British society.

Polling on climate change

Our backgrounds are in environmental psychology, and we have been involved in teams commissioning and analysing surveys of public opinion on climate change. Climate change is an issue that highlights the importance of BAME representation in surveys.

Climate change has a well-established racial justice dimension and people of colour are hardest hit globally. Even in the UK, not everyone suffers equally. Although domestic data is difficult to find, there is clear evidence that many ethnic minority groups are more likely to experience social deprivation, leaving them less able to respond to climate change, are more likely to live in areas exposed to dangerous air pollution, have inadequate access to green spaces and increased risk of overheating.

For all of these reasons, BAME representation in opinion polls is not simply a question of whether people with different coloured skin would answer survey questions differently. The experiences of minority ethnic groups potentially provides a profoundly different set of perspectives, which surveys of public opinion on topics as critical as the climate crisis must aspire to capture.

The first step towards positive change is an acknowledgement that we have not been doing enough to ensure that survey samples are representative of British ethnicity. This is more than just a methodological oversight – it presents a moral challenge to the credibility of many social science surveys.

Want to write?

Write an article and join a growing community of more than 174,500 academics and researchers from 4,804 institutions.

Register now