6.

Some techniques of fertility and mortality estimation

6.1 Background

The way in which a population will evolve, in terms of size and structure, over
time is indicated by the basic demographic parameters of fertility and mortality.
Traditionally, estimation of birth rates and death rates is based on data collected
by census and by a vital registration system. Unfortunately, many developing
countries either do not have a continuous vital registration system or the data
produced by both the system and censuses are of very poor quality that estimates
obtain directly from the data are grossly unreliable.

As mentioned above, population censuses are conventional data sources for the
production of data on the population at risk. Nevertheless, two historical
incidences have created the opportunity to utilize censuses as alternative sources
of data on vital events. These incidences have happened in Canada and the United
States in the late nineteenth and early twentieth centuries, where during those
days the two countries were not having vital registration system nation-wide that
hinders them to have national vital rates, specifically mortality statistics. Hence,
they were in a problem as it is the case in the developing countries of today. It is
the experience of these two countries that triggered the idea of using censuses as
provisional and alternative vital statistics data sources in the currently developing
countries. The attempts and efforts made to use censuses as alternative sources in
the two countries at about the same period, after several tests resulted a non-
promising output that led them to discontinue the method, in Canada in 1911 and
in the United States in 1900. The decision made in both countries at the moment
was to make every effort to establish the vital statistics registration system by
improving and standardizing the procedures of registration of vital events in the
entire nations, in which they have succeeded.

During the 1950th and in the 1960th African countries were facing similar problems
as we have seen above in Canada and the US, even though the two countries have
a well-established census and of course vital registration data in some of the
States. Most African countries during those periods were under the colony
administration, and were not having even a properly conducted census. After
liberation these countries were desperate to have demographic data to design
development plans and programs of their countries. Hence, they were looking for
a quick and shortcut solution that would provide them at least the size and

53
distribution, and growth rate of their population. At this juncture, the alternative
method that was experienced by Canada and the US was introduced as a
provisional solution for the problem facing these countries. These alternative
methods , generally known as 'indirect techniques', were introduced in censuses
and surveys and research has goes on for further refinement of the methods,
which resulted devising numerous analytical methods, which are currently the
major analytical tools for the production of fertility and mortality statistics in the
developing countries.

6.2 Definition and the need for indirect estimation

Estimation of birth and death rates has been based on data collected by census and
by a vital registration system. A continuous registration system usually has the
task of recording vital events (births, deaths, marriages, divorces, etc.) as they
occur. When this system is coupled with periodic counts of the population
(censuses), the calculation of demographic parameters becomes possible.
Assuming that both the registration of vital events and the census counts were
accurate, demographic parameters could be calculated directly from the data
reported and there would be no need for indirect estimation.

Nevertheless, the incompleteness of data and the extent of errors in data collection
necessitated the use of indirect methods of estimation. Some of the possible
deficiencies of a vital registration system, where it exists at all, may be outlined.
The main deficiency is its failure to record all vital events as they occur. For
example, it is well known that births may go unregistered for several years. Only
when the child is ready to join the public education system, or some other type of
organization, for which a birth certificate is required, is the birth reported. In
countries where a sizeable proportion of the population has never attended school
or been in any way connected with official organizations demanding identity
certificates, many births may never be registered. The same may be true of the
births of children who die very young in that respondents may not realize the
need to report the birth of a child who did not survive long enough to have a place
in the household and omit children who died shortly after birth. In such cases, the
parents may consider the registration of either the birth or the death to be futile.
Adult deaths likely either to be reported near the time of their occurrence or not at
all. In most countries, a death must be registered before a burial permit is issued.
Hence, more people may find it necessary to report deaths soon after they occur,
especially in urban areas where burial grounds often restricted to certain areas

54
whose administrators are closely supervised by government officials. However,
the necessity for a burial permit may be practically non-existent in rural areas,
where ties with the government administration are weaker. Therefore, it is not
surprising to find that, in spite of the legal necessity of registration system may
also be deficient in recording characteristics of events, such as age at death, age of
mother at a birth or mother's parity after a birth.

Censuses, the second component of the traditional demographic estimation input,

are also far from yielding perfect data. They suffer mainly from two types of error:
the failure to enumerate all the members of the relevant population (though,
occasionally, some censuses have also produced population over-counts) and poor
age-reporting on the part of the population canvassed. Differential coverage of the
population by age and sex is also very often present (young children, especially,
tend to be undercounted to a greater extent than the adult population), and its
effect cannot always be separated from those of age misstatement.

Because of the impossibility of obtaining reasonable measures directly from the

traditional data sources, demographers have developed a set of techniques that
allow their indirect estimation. The term „indirect‟ used to quantify some of the
techniques used in demographic estimation has its origin in the fact that such
techniques produce estimates of a certain parameter on the basis of information
that is only indirectly related to its value. The classical example is the use of
proportion of children dead among those ever born by women aged 20–24 years to
estimate the probability of dying before age 2. The observed proportion of
children dead is clearly related to mortality conditions, but it is not purely
mortality measure because it is affected by other non-mortality parameters. In
order to transform this proportion into the desired life-table function, the other
parameters must be allowed for, generally by using procedures founded on
demographic models. Therefore, not only is the information used “indirect”, but
also the procedure followed.

The development of these techniques has taken two approaches: either the search
for robust methods to analyze data that have been collected by the traditional
systems or the search for questions that can be answered with reasonable accuracy
and that provide enough information about a certain demographic phenomenon
to permit the indirect estimation of its level. Therefore, indirect techniques are
estimation methods that depend upon models or use consistency checks, or use
conventional data in an unconventional way.

55
6.3 Some indirect techniques of estimating fertility and mortality

The alternative mechanisms used to get the number of vital events and hence the
vital rates, in censuses and surveys are similar. They all try to know the number of
births or deaths by asking retrospectively selected questions either the household
head or the women in the reproductive age. Here, we will limit ourselves to the
type of questions used to elicit the number of vital events in surveys and censuses.
Furthermore, we will also highlight the technique that was develop by extending
the various approaches of fertility and mortality measurements considered in
censuses and surveys, which in general is called indirect techniques for the
estimation of vital rates. However, we will not go to the details of the analytical
tools of the indirect techniques, which are beyond the scope of this course.

6.3.1 Population censuses for collecting data on fertility and mortality

The most commonly used approaches to collect fertility and mortality data
considered here are i) current births and deaths ii) date of the most recent live
birth and iii) children ever born alive and children still surviving. The other
approaches, which we will not consider them here, but exercised by various
countries, are the data that refer to orphan-hood and widowhood methods.

i) Current births and deaths

a) In the case of fertility

In general, current birth information in conjunction with population sizes by sex

and age will enable us to calculate the various fertility measures and reproductive
rates. That are, crude birth rate, general fertility rate, age specific fertility rate, total
fertility rate and also GRR and NRRs.

Births in the last (previous) twelve months

This information enables us to know the total number of births in the year
considered and the data would give the numerator in calculating the
various fertility measures. The presentation or formulating of the question
asked to obtain this information varies from one country to the other. The
question could be addressed to the head of the household or to the woman
herself. The commonly used questions are presented below.

56
 - How many children were born alive in this household in the past year?
(addressed to the head of the household) or
- Was any child born alive in this household last year? (addressed to the
head of the household) or
- How many children did you give birth to were alive last year?
(addressed to women in their child-bearing period, usually 15 to 49
years)

In the 1984 population and housing census of Ethiopia, the question on current
births, which is addressed to the women aged 10–49 was stated as: "Did you give
birth to a baby during the twelve months prior to the census day?" In addition, to
improve the data quality, the reported births were recorded by separating those
alive and those dead in the year. Similarly, the current approach was included in
the 1994 census, where the question was phrased as: "Did ____ have a live birth
during the last twelve months?" Here it was not tried to differentiate whether the
birth(s) were alive or dead. Using the information on the number of births in the
last twelve months and the women in each childbearing age group, age-specific
fertility rates could be derived. An example that shows the distribution of the
births in the twelve months prior to the 1994 census date by the number of women
that enabled to calculate the age-specific fertility rate for the census year is
presented below.

Births in
Age groups No of Women the Past ASFR
Year
15–19 2,937,291 138,529 0.047
20–24 2,166,982 340,499 0.157
25–29 2,006,141 384,470 0.192
30–34 1,600,019 281,340 0.176
35–39 1,458,784 215,524 0.148
40–44 1,153,313 94,960 0.082
45–49 763,543 46,592 0.061

Source: Statistical Report Result at Country Level, Volume I. The 1994 Census,
CSA, Ethiopia.

57
 b) In the case of mortality

As we have seen in the measurement of fertility, here also using current deaths
and the population distribution by sex and age will enable us to calculate crude
death rate, age specific death rate and construction of life tables.

Deaths in the last (previous) twelve months

This information will enable us to collect the total number of deaths in the
year considered and the data would give the numerator in calculating the
various mortality measures. The presentation or formulating of the
question may vary from one country to the other. The question is addressed
to the head of the household. The commonly used questions are presented
below.
 Deaths in the previous 12 months (addressed to the head of the
household) or
 How many deaths have there been in this household during
(calendar year)? or
 Was there any death in this household last year?

For example, in the first Ethiopian population census conducted in 1984, this
current approach was utilized and a question stated as: "Was there any death in
the household during the twelve months prior to the census day?" was presented
to the head of the household. Additional questions, such as age and sex of the
deceased were included.

The table (Table 4.4, page 44) that shows the age specific death rate that is used in
the construction of the life table for rural Ethiopia, in chapter four is a good
example of the data outcome from the current death data collection approach.

ii) Date of the most recent live birth

This approach is introduced to acquire an improved data on recent fertility and

infant mortality. It included collecting the date of birth of the most recent child
born alive and the survival status of this child at the time of the census or survey
inquiry. This approach was meant to minimize the well-known problem of
misunderstanding of reference period. This approach was not introduced in the
two Population and Housing Censuses of Ethiopia and most surveys conducted
by the Central Statistics Authority.

58
 iii) Children ever born alive and children still surviving

The information on children ever born alive will give us the level of fertility in a
particular population. In this approach the question(s) will be addressed to all
women in the reproductive ages, which will provide the number of children the
women had up to their current age. The number of children obtained from each
woman will be aggregated and classified by the number of women in the
reproductive age groups, which is used to calculate the cumulative fertility at each
age group. The approach that uses children ever born to obtain fertility level is
known by various names, that is, lifetime fertility, or completed fertility or average
parity. Originally, the approach uses a single question, such as, "How many
children have you ever borne alive?" This simple question has been extensively
used, but examination of the results it yields suggests that children are sometimes
omitted, particularly by women aged 35 and over. Children who have left home or
who have died are likely to be omitted. It is, therefore, recommended that when
the time allotted to each interview and the questionnaire space permit, the basic
question be broken down into three parts. Accordingly, the question was modified
to improve the quality of the data and it is decomposed into three components as
follows:

Of the children you have ever borne alive how many:

1. are still living at home with you?
2. are still living, but elsewhere, in some other household?
3. have died up to the census/survey date?

The application of this approach to provide mortality estimate was developed

later, by sharing the data obtained from the modified questions, that is, the three
separate questions. Hence, from children ever born alive approach it is possible to
make mortality estimates using analytical methods. That is, applying the
information on children ever born and children surviving it is possible to calculate
mortality levels for various age and sex groups, using the indirect methods of
estimation.

In both the 1984 and 1994 censuses the children ever born alive and children
surviving data collection approach was employed to calculate fertility and
mortality measures. The tabulated data of the 1994 census that are obtained using
the children ever born approach are presented below.

59
 Children Mean
Children
Age groups No of Women Ever Born Children
Surviving
Alive Ever Born
10–14 3,619,206 17,145 15,695 –
15–19 2,928,890 583,897 515,375 0.2
20–24 2,158,164 2,736,722 2,369,471 1.3
25–29 2,003,341 5,705,196 4,817,984 2.8
30–34 1,598,491 7,093,351 5,825,559 4.4
35–39 1,457,685 8,001,424 6,462,459 5.5
40–44 1,151,931 6,967,360 5,440,330 6.0
45–49 762,440 4,745,920 3,634,233 6.2
50–54 782,502 4,640,494 3,418,975 5.9
55+ 1,593,649 8,495,050 5,836,776 5.3

Source: Statistical Report Result at Country Level. Volume I. The 1994 Census,
CSA, Ethiopia.

6.3.2 Household sample surveys for collecting data on fertility and mortality

As we have seen above for censuses, surveys are one major source of birth and
death statistics for countries with incomplete vital statistics registration system. By
their nature censuses are conducted every ten years. Hence, census would not give
birth and death rates for the inter-censal period, which requires using sample
surveys at least to have data for some of the years in the inter-censal period.
Household sample surveys could vary in their types and complexity, but they
could broadly be categorized as, single-round sample surveys, multi-round
(follow-up) surveys and dual-records system. The basic difference between them
lies in the time reference and measurement approach. For instance, with respect to
time reference the difference between the first two survey types is obvious, as
their names indicate the first is a one time study while the second is continuous
either on a quarterly or half yearly or yearly basis. The difference with the third
method goes beyond the time reference, where the dual-records system even
though, in principle it should be done on a continuous basis, however, its
measurement method is different from the others. In the first place the dual-
records system was basically developed to provide a better and qualified vital

60
rates, secondly, the system uses two independent data collection methods, that is,
a periodic household survey and a separate vital event reporting method.

In general, the various alternative current and retrospective questions that apply
to obtain fertility and mortality estimates, and that were discussed in their
application to censuses; are the once that are used in single and multi-round
surveys. Specifically, the single-round demographic surveys use similar questions
and approaches in collecting and calculation of vital rates as we have seen in
census, except that surveys will use their own population sizes (that is obtained
from the survey) as input for the denominator, that is, the population at risk. In
addition to the characteristics referred to single round surveys, multi-round
surveys have various advantages, specifically in reducing the memory-lapse, by
interviewing households repeatedly within shorter time when compared to single-
round surveys and census.

6.3.3 Indirect techniques for the estimation of birth and death rates

The data obtained from the current and retrospective approaches employed in
censuses and surveys could not be taken as true values. There are various types of
errors incurred during data collection by the data collector as well as by the
respondents. In addition to the data errors due to the data collectors and
respondents, the nature of the questions is one source of the inherent problem of
the approaches. Hence, the quality of the information obtained through these
current and retrospective methods is found to be poor which requirs further
refinement and adjustment. For instance, from what has been observed in the
various census and survey taking activities, the number of births or deaths
reported in the twelve months before the census or survey date is always
underreported. Similarly, in the case of reporting of children ever born alive to
women in the childbearing age, it was observed specifically, women at the higher
ages are more likely to forget their children born alive, which is revealed when the
number of children alive is classified by the number of women in each age group.

The major factor for the poor performance of these approaches in general is recall
error or memory lapse. The other important sources of error are misdating of the
reference period of the birth or death, misstatement of age of women, erroneous
inclusion of fetal deaths, or the inclusion of adopted children or on the contrary,
total omission of the birth or death.

61
In general, the indirect techniques of demographic estimation methods utilize data
collected through surveys and censuses on current and retrospective approaches,
that is, current births and deaths, children ever born alive, children surviving, date
of most recent child born alive and its survival status, etc., to estimate various
fertility rates and infant and child mortality rates. For instance, in the case of
mortality, using data on children ever born alive and children surviving, infant
and child mortality rates could be estimated while estimation of adult mortality
could be made using data obtained through survival of mother and father, or
survival of the first spouse data approaches. Thus, the calculation of the various
fertility and mortality measures using these methods is highly dependent on the
assumptions and mathematical models used, which determines the validity of the
results (for details see UN, 1983).

Reference

UN (1983). Indirect Techniques of Demographic Estimation. Manual X, Population Studies No. 81.
Department of International Economic and Social Affairs, United Nations. New York.

62