Comparing ethnicity on death registrations with ethnicity on birth registrations
The ethnicities that are recorded on a person’s death registration do not always match the ethnicities that are recorded on their birth registration. This makes it harder to compare the death rates of different ethnic groups. The article also presents data on comparing the two registrations and discusses the implications for estimating death rates.
The recording of ethnicity on birth and death registrations
Birth and death registration forms have questions about ethnicity. Both forms allow respondents to record more than one ethnicity. Birth registrations are usually filled out by the child’s parents, but death registrations are usually filled out by the funeral director. Each ethnicity on the birth or death registration is coded to the most detailed level of the New Zealand standard classification. However, birth and death data, and the population estimates, projections, and rates derived from them, are generally published for broader ethnic groups, such as ‘Asian’, ‘European’, ‘Māori’, ‘Middle East, Latin American, or African’, ‘Pacific’, and ‘Not Specified’.
Trends in reported ethnicities
Figure 1 shows the average number of ethnicities recorded on children’s death registrations, and the number recorded on the same children’s birth registrations. The data are for children who died before their fifth birthday, beginning in 1998, the first year for which electronic data are available, and finishing in 2011. The graph shows, for instance, that children who died in 1998 had an average of about 1.1 ethnicities recorded on their death registration form and 1.2 on their birth registration.
The average number of ethnicities reported on death registrations is consistently lower than the number on birth registrations. The reasons for the difference are not clear. However, most analysts assume that parents’ descriptions of their children’s ethnicities are more accurate than those of funeral directors.
Average numbers of ethnicities on birth and death registrations have been trending upwards since 1998. This is consistent with census data, which also show the average number of ethnicities to have been rising.
A downloadable data file from 'Available files' gives the underlying numbers from figure 1.
Comparison of birth and death registrations
How closely do the ethnicities recorded on birth registrations align with the ethnicities recorded on death registrations? For instance, if a child’s birth registration says that the child is ‘European’ only, what is the death registration likely to say?
Figure 2 summarises data on the alignment of birth and death registrations for the main combinations of ethnicity. The results were calculated from all birth and death registrations for the period 1998–2011 for children who died before their fifth birthday where the records could be ‘matched’. Registrations were ‘matched’ if they clearly referred to the same person.
The widths of the blocks in figure 2 are proportional to the number of death registrations. For instance, the left-most block in the top row shows, for children who have ‘European' only on their birth registration, what proportion has ‘European' only on their death registration. This block is by far the widest in the first row because most children with ‘European’ only on their birth registration have the same on their death registration. (The exact figure is 88 percent.) The ‘European and Māori’ block is the second-widest block in the first row, because ‘European and Māori’ is the second most common ethnicity on death registrations for children who have ‘European' only on their birth registrations.
The heights of the blocks are proportional to the number of birth registrations. The blocks for ‘Māori' only, for instance, are the second highest, because this is the second-most common ethnicity on the birth registrations.
Implications for estimating death rates
Death rates for an ethnic group are calculated by dividing (i) the number of deaths of people in the ethnic group by (ii) the number of people in the ethnic group. Data on (i) come from death registrations. For children under one year of age, who comprise 85 percent of the records analysed here, data typically come from birth registrations. When ethnicity is recorded differently on birth and death registrations, estimates of death rates using reported ethnicities can be misleading, unless counter measures are taken.
One such counter measure is to match birth registrations to death registrations, and to use only the ethnicities recorded on the birth registrations. This maintains consistency between births and deaths, and also between children who die and children who survive. The New Zealand Census Mortality Study follows an equivalent strategy with census and deaths data.
Implications for the use of administrative data
Birth and death registrations are examples of ‘administrative data’, that is, data generated through general purpose administrative procedures rather than a specialised survey. Administrative data hold great promise as a way of producing detailed population and social statistics at low cost. But, as the inconsistencies in ethnic reporting illustrate, administrative data need special care. The responses that are obtained through one administrative procedure do not necessarily match the responses obtained through another procedure. The accuracy of the responses in an administrative dataset depend on many things, including the identity of the person providing the data, the wording of the question, and the circumstances under which the question is asked.
Technical details about the data and figures
The data presented in figures 1 and 2 come from a dataset assembled by Statistics New Zealand. The dataset contains birth registrations since 1998, the year electronic recording of birth and death registrations began. It also includes death registrations for children born since 1998. Wherever possible, each death registration has been linked to the individual’s birth registration. The linking was done using first name, last name, date of birth, and meshblock of birth. The linking was done using specially-written computer programs, supplemented by manual searches. If birth registrations and death registrations had minor inconsistencies, such as slightly different spellings of names, they were still matched, provided that the weight of evidence suggested they referred to the same person.
Figures 1 and 2 refer to children aged 0–4 years at the time of their death. There are 4,560 such cases in the dataset. Among the 4,560 cases, 85 percent of death registrations have been matched with birth registrations. The match rate varies by ethnicity, from 94 percent for people with ‘European' only ethnicity to 75 percent for people with ‘Māori' only. The reasons for the different match rates are not clear. Unmatched cases are omitted from figure 2. Ideally, comparisons of death rates should take into account the extra uncertainty created by failures in the matching process. Statistics NZ is investigating methods for doing so.
Beginning in August 2013, the Department of Internal Affairs have included a birth registration reference number in the death registration dataset provided to us each month. This will make it easier to carry out future comparisons between the characteristics recorded at birth and death.
Figure 2 is a mosaic plot, created using the function mosaic with the statistical package R.
For more information contact:
Christchurch 03 964 8700
For general enquiries contact our Information Centre:
Phone: 0508 525 525 (toll-free in New Zealand)
+64 4 931 4600 (outside New Zealand)
ISBN: 978-0-478-42914-5 (online)
Published 6 August 2014