Ethnicity of the deceased person: further background

Estimating death rates by ethnic group, age and sex

The 2011 Census breakdown of the population of Scotland by ethnic group, age and sex was published in February 2014. It was then possible to estimate death rates by ethnic group, age group and sex, using the number of deaths registered in 2012 as the numerator and the population size on 27 March 2011 as the denominator. When National Records of Scotland (NRS) did so, without publishing the results, it noted that there was some uncertainty about their reliability because, for example:

• non-White numbers would more than double if, say, a fifth of the ‘ethnicity not known’ deaths were non-White;
• the deaths included people who were not counted in the 2011 Census (for example because they were living outwith Scotland at the time). However, including such deaths (for example of visitors to Scotland) ‘offsets’, to an unknown extent, not counting the deaths elsewhere of people who were residents of Scotland at the time of the 2011 Census, but who subsequently left (or died while out of the country);
• the age-groups have different bases – deaths are counted on the basis of the age at the time of death, whereas the 2011 Census population is on the basis of the age on the census date. For example, someone who was 64 at the time of the 2011 Census might die aged 65 in 2012 (or 66 in 2013 or …) – so the figures for deaths aged (say) 60-64 in 2012 will exclude the deaths of some people who were aged 60-64 at the time of the 2011 Census, and will include the deaths of some people who were under 60 at the time of the 2011 Census.
• rates based on small numbers may not be reliable, because they may be subject to large percentage year-to-year fluctuations. For example, ‘only’ 46 deaths of non-White people aged 85+ were registered in 2012: 23 males and 23 females.

Therefore, although NRS had published both the numbers of deaths registered in 2012 and the 2011 Census population, broken down by ethnic group, age-group and sex, NRS did not publish the death rates that it had estimated from those figures.

The first public presentation of NRS’s estimates of death rates by ethnic group, age and sex

At a Migrant and Ethnic and Health Research conference in Glasgow in September 2015, an NRS Vital Events Statistician presented a summary of the results of the first three years’ collection of the data (2012, 2013 and 2014). The full presentation can be found on The statistics that have been published page of this website. This included the following figures:

• 163,876 deaths were registered from January 2012 to December 2014;
• an ethnic group was recorded in 157,232 cases (95.9 per cent);
• the ethnic group was a ‘White’ one in 156,309 cases (95.4 per cent) and a ‘non-White’ one in 923 cases (0.6 per cent);
• the informant was not willing to provide the information in 5,880 cases (3.6 per cent); and
• the informant was unable to provide the information, because he/she did not know the deceased’s ethnic group, in 764 cases (0.5 per cent). 

The results also included some estimates of apparent death rates by ethnic group, age and sex, produced using the numbers of deaths registered in the three years and the population figures from the 2011 Census. (The dates of birth from the death registration records were used to determine how old each person had been on the Census date, 27 March 2011. The age-groups used were 70-74, 75-79, 80-84 and 85+. Figures for younger age-groups were not given because they would be based on smaller numbers of deaths.) The Statistician emphasised that such death rates would have to be used carefully, because the ‘non-White’ numbers would more than double if, say, a seventh of the ‘not willing / unable to provide’ deaths were actually deaths of non-White people. Might a ‘not willing to provide’ or ‘not known’ response be more likely in the case of the death of a non-White person? Also, rates calculated from relatively small numbers of deaths could have large percentage period-to-period fluctuations. For example, for non-White people who were aged 70-74 at the time of the 2011 Census, there were annual averages of ‘only’ about 30 male and 18 female deaths per year over the three years. Another potential complication was the fact that the population at risk (this is the number of people in Scotland, broken down by ethnic group, age and sex) might have changed since the 2011 Census. So the results that had been produced might be unreliable.

When NRS estimated death rates by ethnic group, age and sex in this way, it appeared that, as would be expected:

• death rates increased with age (this was the case for the death rates for White males, White females, non-White males and non-White females) and
• death rates were generally higher for males than for females (this was the case for Whites for all four of those age-groups, and for non-Whites for the first three age-groups).

However:

• the apparent death rate for non-White males aged 85+ was only slightly higher than the apparent death rate for non-White males aged 80-84; and
• for non-Whites aged 85+ (at the time of the 2011 Census), the apparent death rate was lower for males than for females.

The Statistician said that the last two points raised questions about the reliability of the results. The apparent death rate for non-White males aged 85+ was low, when compared with the estimated death rates for others aged 85+, and with the level that might be expected based on the normal pattern of death rates by age-group and sex.

The Statistician referred to some intended future work. In due course, NRS would try to match each death record to that person’s census record (if the deceased was in Scotland on the date of the 2011 Census), in order to be able to compare the ethnic groups in the two records to observe (for example) if they were the same, and – in the case of deaths for which no ethnic group had been recorded – what ethnic group (if any) had been reported in the census. NRS would then be able to assess the quality of the ‘ethnicity of the deceased’ data (for example was there significant under-recording for some groups?) and the implications for death rates that were calculated using those data.