Limitations of ONS Life Expectancy Estimates for minoritised ethnic groups in the UK

Harry Taylor, Dharmi Kapadia & Laia Bécares

Centre on the Dynamics of Ethnicity (CoDE), The University of Manchester

There is a wealth of evidence showing that people from most minoritised ethnic groups have much poorer health than White British people. These studies have shown that not only do ethnic inequalities in health exist, but that they have persisted over time, and are exacerbated in later life.

It is therefore surprising that life expectancy estimates from the Office for National Statistics (ONS) published in 2021 show that many minoritised ethnic groups in England and Wales have very high life expectancies, often higher than White British people, despite having some of the poorest health outcomes throughout their lifetime. How can we explain these opposing findings?

The ONS has been clear that these estimates are experimental; however, we have observed these estimates being cited without caveat or any discussion of their experimental nature in academic research, local government briefings, and government policy documents. This is worrying, as if questions are not raised over their accuracy, these estimates may continue to be used in policy discussions, service planning, and public debate, potentially resulting in a lack of funds and resources to tackle ethnic inequalities in mortality.

We examined the methodology, data and assumptions used by the ONS in constructing their estimates of life expectancy by ethnic group. We found that missing data were an important factor, and that assumptions on how to overcome the problem of these missing data had a large impact on the life expectancy outcomes.

The ONS approach relies on linking people recorded in the Census to subsequent death registrations. There are several obstacles in implementing this linkage. First, we may be unable to tell if someone has left the country and second, we cannot reliably observe what has happened to people who have left the country. Life expectancy calculations depend on two basic components: the size of the population at risk, and the number of deaths within that population. If either of these figures are inaccurate, the life expectancy estimates will be flawed.

Further, our research found that knowing whether some is still alive or has died is not evenly distributed across ethnic groups. In longitudinal analysis looking across the decade between censuses, around 9% of White British people are not recorded at follow-up. For minoritised groups, the proportion missing is substantially higher, reaching around 30% for some groups. We believe that many of these people may have emigrated, yet remain in the population base. As their death will never be recorded in UK administrative statistics, they become “statistically immortal”. This leads to an over-estimation of life expectancy for the ethnic minority groups affected by this.

The lack of visibility of life outcomes of people who emigrate may also affect minoritised ethnic groups disproportionately. Research in the field has suggested that migrants who are unwell return to their country of origin at the end of their lives. However, a study by Guillot and colleagues looking at the mortality of French pensioners found that a person born outside of France, who then migrates out of France is between 1.6 and 3.5 times more likely to die than a person born outside of France, who stays in France. This research used pensions data to create a marker of whether individuals were still alive after they emigrated, finding that people who emigrated did not live longer. Although the ONS is working on admin-based datasets that would produce similar markers of economic activity and allow us to do this kind of research on UK data, these are not currently available to researchers.

In our research, we estimated what might happen if we applied the mortality risks observed by Guillot and colleagues to the UK context. We substituted the ONS’ assumptions on the quantity of outmigration, with our own estimates using data from the ONS Longitudinal Study, and substituted assumptions on the mortality rate of people who emigrate with those produced in the research by Guillot and colleagues. We then demonstrated how the life expectancy varied for each ethnic group according to these assumptions.

The purpose of our research was not to produce definitive estimates of life expectancy (which is not possible to do with the available data), but rather to show how much variability there is in the ONS’ life expectancy estimates when we use alternative methodologies, and give full consideration to the unknown factors in the available data.

What we found is that the life expectancy estimates for minoritised ethnic groups are highly unstable, yet this instability is not mirrored to the same degree for White British people, because the underlying missingness problem is not as severe. This illuminates the risk in uncritically referencing the ONS’ experimental statistics, and presenting them as definitive rather than experimental.

We advise researchers and policymakers to treat the ONS’ life expectancy estimates with caution, and focus on what is already well-established: substantial ethnic inequalities in health exist across the life course, they persist over time, and they often worsen in later life.

Work to ensure that the analysis of ethnicity data is used to tackle racial health inequalities, rather than entrench them will be a core theme at the forthcoming conference, ‘Advancing Racial Justice: Convention on the State of Data 2026’, where we will explore how to build data systems that illuminate inequality, and how to ensure that ethnicity data serves justice.

Taking place on 16 April 2026, the conference will create a unique, cross-sector space for agenda-setting on how data can drive meaningful action on racial inequality. Registration is now open, and we are inviting those working on data, equity, public policy, public sector or community outcomes to join us.