Professional Documents
Culture Documents
M. G. SANTOW
National Blood Pressure Study, P.O. Box 691, Canberra City, A.C.T.
Received
I December
2601, Australia
ABSTRACT
A microsimulation model is used to assess the extent to which the fertility of the
Yoruba of Western Nigeria may be affected by changes in the durations of lactation and
marital sexual abstinence. The simulations make no allowance for the compensatory use
of contraception. A series of preliminary simulations demonstrate the effect on fertility of
the length of the period of post partum non-susceptibility to conception, and an attempt is
made to duplicate the reported fertility of a large Ibadan survey (CAFNl). Input data are
then drawn from a number of recent Nigerian demographic surveys which enable the
separate simulation of the fertility of rural dwellers, poorer Ibadan women and richer
Ibadan women. The output indicates that, in the absence of contraception, urban fertility
is likely to exceed rural, and the fertility of richer urban women is likely to exceed that of
poorer urban women.
1.
INTRODUCTION
BIOSCIENCES
BElsevier North-Holland,
Inc., 1978
93
0025-5564/78/090087 + 25.SO2.25
M. G. SANTOW
94
The last Nigerian census was taken in 1963 and produced a sizeable
overcount (Lucas and Williams [ 131). More importantly,
the establishment
Caldwell and Caldwell [6] write that the explanation for [the effect of a new
conception on the health of the unweaned child]. . . is not always the correct one but the
incorrect explanation is often more vivid and probably more efficacious in enforcing
abstinence; amongst the Yoruba it is widely believed that the mans sperm actually enters
and poisons the milk which is being fed to the baby.
MICROSIMULATION
OF YORUBA
FERTILITY
95
The second problem, that of age misstatement, may cause the misclassification of reported vital events and therefore distort age-specific fertility and
mortality rates (see van de Walle [25]). All African demographic surveys
share the problem of trying to record the ages of people who do not know
their exact ages and are not fundamentally
interested in knowing them
(van de Walle [26]). The most common source of error in all age reporting is
the overstatement
of ages ending in certain preferred digits, with a corresponding understatement
of ages ending in other digits (Nagi, Stockwell
and Snavley [ 151). We shall return to this topic at a later stage of the paper.
Microsimulation
provides a useful technique for measuring the effect on
fertility of rapid change amongst a group of intermediate variables, in this
case breastfeeding and post partum and terminal abstinence. For the sake
of clarity several areas of change were ignored in the simulations, namely,
infant and child mortality
and contraceptive
usage. Firstly, a recent
Nigerian study has demonstrated a very real differential between the infant
and child mortalities of two villages which differed principally
in their
access to public health services (Orubuloye and Caldwell [19]). The implementation
of new health services is continuing and may be expected to
lead to an increase in the age-specific proportions of surviving children, as
the greatest effect of such services is the reduction of the exogenous
component of infant and child mortality. Secondly, although the level of
contraceptive
use is low amongst the Yoruba of Ibadan, it has been
21n a Liberian study, Gay and Cole [8] observed
even domestic
d/
MICROSIMULATION
/\
OF YORUBA FERTILITY
-1:;
97
98
M. G. SANTOW
increasing
over the last years (Caldwell and Caldwell [5]). The use of
microsimulation
is indicated
in this instance not only because of the
unreliability
of the retrospective
fertility data at our disposal, but as a
means by which to identify the effect on fertility of changes within one
group of linked intermediate variables in isolation from a number of others.
3.
THE MODEL
The microsimulation
model used in the following analysis is based on
those pioneered by Hyrenius and Adolfsson [lo] and Perrin and Sheps [20].
The time unit of the model is the so-called lunar month of 28 days, as this
is, on average, the longest period of time during which no more than one
conception can occur (Santow [23]). The reproductive
span of each simulated woman is defined by her ages at marriage and final sterility. Thus no
allowance is made for the possibility of pre-marital conception or marital
dissolution by divorce or the death of either spouse. As the Yoruba woman
generally remarries soon after her widowhood,
the assumption
that she
remains in some sort of sexual union for as long as she is fecund is not
unwarranted.
The male partner does not appear explicitly in the model, but,
where appropriate, his contribution
to fertility is represented by means of a
couple parameter.
For example,
the probability
that a conception
terminates in a spontaneous
abortion is a female parameter. On the other
hand, the probability
that a conception
occurred in the first place is a
couple parameter, because it is dependent on both of the marriage partners.
The first step in the simulation
of each reproductive
history is the
determination
of the womans ages at marriage and final sterility. (Sterility
data are generally derived from the fertility non-experience
of couples.)
The Monte Carlo technique is applied to the two relevant input distributions, and if the age at which sterility occurs does not exceed the age of
marriage, the woman exits from the simulation and the process is repeated
on the next woman.
Given that a positive number of reproductive cycles is initially allocated
to the woman under simulation,
the Monte Carlo method is used to
determine whether she will conceive in the first cycle. If the first test is
unsuccessful it is repeated, and the number of such trials before a success
occurs gives the waiting time to conception in lunar months. The value of
the random number which identifies a conception also indicates whether the
conception is to terminate in a spontaneous
abortion, a stillbirth or a live
birth. The lengths of the periods of gestation
and of post partum
amenorrhea
or sexual abstinence are determined from the input distributions specific to the pregnancy outcome.
The lunar-month
counter is set to zero at marriage. After each Monte
Carlo test this counter is incremented
either by one, in the case of an
MICROSIMULATION
99
OF YORUBA FERTILITY
unprotected
cycle in which no conception occurs, or by the duration of
non-susceptibility
after a conception. This latter term is calculated as the
sum of the durations of pregnancy and post partum amenorrhea or sexual
abstinence. After each such incrementation
the time counter is compared
with the predetermined
reproductive
span, and as soon as the counter
exceeds this span the reproductive history is terminated. Its fertility data are
stored, and the next reproductive history is simulated. When the sample is
complete the final aggregates are made. All the simulations are of 1000
women.
4.
INPUT
DATA
The input data are of two types. The first comprises biological input not
drawn specifically from the Yoruba but compatible, as far as is known, with
their condition. The second consists of data drawn from a number of recent
Nigerian demographic surveys. The following examples demonstrate
how
the input data are inserted into the model.
We wish to determine whether a non-pregnant,
fecund woman of a
particular age will conceive. In this example she is 22 years old, and thus the
probability of her conceiving is 0.20. A random number is selected, and
according to whether it is less than or exceeds 0.20, a conception is said to
occur or not. In the former case the random number also determines
whether the conception will terminate in an abortion, a stillbirth or a live
birth. A second random number is selected and is compared with the
appropriate
cumulative frequency distribution
from Table 1 in order to
TABLE 1
Percentage Distributions of Duration of Gestation and Post Partum Amenorrheaa
Gestation
Months
1
2
3
4
5
6
I
8
9
10
11
12
Mean
Abortion
Stillbirth
Abortion
40
60
21
19
8
3
2
1
30
10
15
19
30
36
2.2
9.9
1
4
44
48
3
10.5
1.5
Stillbirth
4
49
31
12
3
1
2.6
100
M. G. SANTOW
BIOLOGICAL
DA TA
The fecundability
input is a simple age-dependent
function increasing
linearly from a value of 0.13 at age 15 to 0.20 at age 20, remaining constant
until age 25 and then declining linearly to a value of 0.03 at age 42, at which
value it remains constant (Santow [23]). Spontaneous abortions account for
15 per cent of all conceptions, and stillbirths for a further 2 per cent.
The sterility input is an exponential function derived from Pittenger [21]
with a radix of 5 per cent of women sterile at age 17.5 and a median of 41
years.3 Whereas fecundability
is determined solely as a function of age and
is the same for all fecund women of the same age, the point at which
sterility overtakes each woman is determined by the Monte Carlo method.
4.2.
SURVEY
DATA
MICROSIMULATION
OF YORUBA
FERTILITY
101
nicely fulfilled by the reported age statistics of the CAFNl women. There is
massive heaping on ages terminating in a 5 or a 0 (particularly between the
highly fecund ages of 20 and 30) although more women reported that they
were 25 than reported that they were 20. Preference was then given rather to
even terminal digits than odd. More specifically, the digits 2 and 8 were
more popular than 4 and 6, since these latter digits are overshadowed by the
highly preferred 5. Similarly, the digits 3 and 7 were less unpopular than 1
and 9.
The eccentricities of this population pyramid cast doubi not only on
reported age-specific fertility data but also on the reported distributions of
age at marriage. After smoothing, the means of the distributions
obtained
from the NF2-1, NF2-2 and NF2-3 surveys were respectively 22, 18 and 22
years, and their ranges were (15,26), (14,21) and (16,28). The mean age at
marriage of the CAFN 1 distribution
was 21 years, and the range was
(17,26).
Figure 3 presents graphically the Nigerian Family Project distributions
of the length of lactation, the corresponding length of amenorrhea and the
length of post partum abstinence. The duration of amenorrhea
was not
sought in any questionnaire
but was estimated using data from studies
which link lactation and post partum amenorrhea.5 The graphs facilitate
such comparisons as the percentage of women in each sample who are no
longer effectively sterile on their babys first birthday. Through abstinence
this percentage increases from 7 (NF2-1) to 21 (NF2-2) to 74 (NF2-3), while
through breastfeeding,
the percentage increases from 46 (NF2-1) to 56
(NF2-2) to 100 (NF2-3). The figure shows quite clearly that for the richer
women (NF2-3) it is nursing, rather than abstinence, which provides the
greatest contraceptive protection for the first six months after confinement.
In other words, although the erosion of traditional practices has affected the
lengths of time both for which women are willing to breastfeed and for
which they are willing to abstain, it is the custom of post partum abstinence
which has been affected, in this case, the more drastically.
The lowest strip on the graph compares three pairs of distributions which
were obtained by asking women both how long post-natal
abstinence
should continue, and how long they themselves had abstained after their
last live birth. Two pairs are drawn from the Fertility and Family Limitation Survey (FFL) and the third, represented by unjoined points, from the
large Ibadan survey (CAFNl).
In both sets of Ibadan data the should
abstain distribution lies fairly consistently above the did abstain distribution. One explanation for this is that the respondent does not anticipate
the death of an unweaned baby when she answers the should abstain
%e
Santow
of such studies.
NF3-4
NF3-3
NF3-2
NF3-1
NF2-3
NF2-2
NF2-1
Data
identification
The Nigerian
FamilyProject 3,
Terminal
Female
Sexual
Abstinencea
The Nigerian
FamilyProject 2,
Post-natal
Female
Sexual
Abstinence
Data source
19745
19765
Date of
field work
Western State
Western State
Area
Yoruba
women
aged
Yoruba women
who had borne
at least one
child who had
survived
until weaning
and
resumption of
sexual relations
Respondents
140
Poorer Ibadan
Richer Ibadan
70
Town
Village
Richer fbadan
Poorer fbadan
Additional
characteristics
70
I 140
Size
Sample
TABLE 2
Characteristics of Different Sources of Nigerian Data Used in Simulations Incorporating Post Partum Sexual Abstinence
Fertility and
Family
Limitation
surveyb
FFL-1
1973
Ibadan City
Yoruba
women aged 15-59
6606
(Sampling
fraction
=1:24)
-- -
- -
.- ^
- .- .-_ ., _^ _
^ _ -
> .> -
, _ ,.r, .. .- -
_ .- --
%rveys based on the Sociology Department of the University of Ibadan. The Changing African Family Projects were
joint undertakings of the Sociology Department of the University of Ibadan and the Demography Department of the
Australian National University and were funded by the Population Council. The Nigerian Family projects were
undertakings of the Demography Department of the Australian National University and were funded by the Department.
bSurvey conducted by I. 0. Orubuloye, currently a Ph.D. student in the Demography Department of the Australian
National University-survey
funded by the Demography Department.
FFL-2
The Changing
African Family
Project: Nigerian
SegmentProject 1,
The Beginning
of Family
Limitation
CAFNl
M. G. SANTOW
104
2.i5
20 -
POPULATION
FIG. 2.
LN
HUNDREDS
MICROSIMULATION
OF YORUBA FERTILITY
M. G. SANTOW
106
analogous with that observed in the reported age statistics, as the durations
of abstinence and nursing are often linked directly to the childs age. For
instance, a woman may resume sexual relations during a particular festival
at harvest time, recalling that she gave birth during this festival two years
before.6 The data presented in Figure 3 provide input distributions
of post
partum amenorrhea and sexual abstinence specific to each survey, using a
time unit of half a year.
In the CAFNl
sample, 55 per cent of women aged 40-44 years had
terminated all sexual relations, while 69 per cent had done so at ages 45-49,
and 83 per cent at ages 5&54. A further survey (NF3) was devoted
exclusively to female terminal abstinence and provided clear differentials in
the ages at which rural, poorer and richer Ibadan women became terminally
abstinent. By the age of 44 years the number of terminally abstinent women
comprised 53 per cent of the rural women (NF3-1 and NF3-2), 30 per cent
of the poorer Ibadan sample (NF3-3) and 30 per cent of the richer Ibadan
women (NF3-4). The average ages of commencement
of terminal abstinence varied only from 36.4 years @F3-3) to 40.7 years (NF3-l), but the
age distributions
of entrance into the abstinent state showed considerable
variation.
The NF3-1 and NF3-2 data were aggregated to provide an age distribution of entry into the terminally abstinent
state compatible with NF2-1
data. Similarly, the NF3-3 terminal abstinence
data were applied to the
NF2-2 simulation, and the NF3-4 data to NF2-3. Single-year distributions
were constructed, and as the NF3 sample included no women older than 44
years, the proportions
of terminally abstinent women at higher ages were
obtained by linear extrapolation,
with all women terminally abstinent by
the age of 50.
5.
FERTILITY
AND THE PERIOD
CEPTIBILITY
TO CONCEPTION
OF POST PARTUM
NON-SUS-
MICROSIMULATION
107
OF YORUBA FERTILITY
TABLE 3
Average Parity by Duration of Post Partum Non-susceptibility
Non-susceptibility
(lunar months)
Average parity
Width of 95%
confidence
intervala
to Conception
13
18
25
35
12.94
11.52
9.92
8.66
7.55
6.56
5.32
0.54
0.47
0.42
0.38
0.31
0.27
0.23
aThe width of the 95% confidence interval is obtained by dividing twice the sample
standard deviation by the square root of the sample size, and multiplying the result by
1.96.
L
FIG. 4.
M. G. SANTOW
108
then
&-1-s.
(2)
Consequently
it is this factor, rather than the use of a constant duration of
post partum non-susceptibility,
which is responsible for the shrinking of the
confidence intervals.
6.
PRELIMINARY
TESTING-CAFN
1 DATA
The next task assigned to the model was to attempt to simulate the
fertility of the CAFNl
sample. This survey sought no information
on
breastfeeding,
but we have seen that the proportion of women still breastfeeding exceeds the proportion
still abstaining
only during the first six
months post partum of the richer Ibadan women (NF2-3). Caldwell and
Caldwell [6] noted that
in rural areas, the period of abstinence
is shorter than the period of lactation even now in
only about one case in twenty where the child survives, and further investigation
usually
reveals that such atypical
behaviour
is exhibited
only by persons who have broken
substantially
with the traditional
culture.
in all but the most exceptional
cases it is abstinence, rather
than post partum amenorrhea,
which provides the greatest post partum
protection against conception, as the duration of lactation exceeds that of
post partum amenorrhea.
Two simulations were performed which incorporated
the CAFNl did
abstain and should abstain distributions,
and the CAFNl
estimated
distribution of female age at marriage. The 95 per cent confidence intervals
around the average completed fertilities of these runs were, respectively,
(5.14,5.42) and (5.25,5.54), and their means were 5.28 and 5.40 live births.
Consequently,
MICROSIMULATION
OF YORUBA
FERTILITY
109
*Contraception here includes the use of charms or medicines sold by a native doctor,
rhythm, withdrawal, condoms, jellies, creams, sQline pessaries, douching, diaphragm,
foam, internal ring, orals, I.U.D., the sterilization of either partner and abortion. See
Caldwell and Caldwell [5].
M. G. SANTOW
FIG. 5. Distribution of live births of real and simulated populations with long periods
of post partum sexual abstinence.
1,
then
We assume that ni+ = nip =0 when i is even, and that the extent of
individual misreporting does not exceed one birth. Then the reported mean
fertility, FR, can be expressed as
(4)
111
Consequently,
FR = F when
(5)
Thus, if the total number of odd-parity women who report one extra birth
equals the number who report one fewer, that is, if the direction of the
distortion is random, then the average reported fertility will still represent
the average achieved fertility.
A stronger condition for equality is that
nZ+
7.
I=
BREASTFEEDING,
DATA
nzi+3
for
i=O, l,..,k-
ABSTINENCE
1.
AND
(6)
FERTILITY-NF2
The sizes of the NF2 sub-samples preclude the direct estimation even of
average completed fertility. On the other hand, the distributions
of post
partum non-susceptibility
to conception are no more irregular than that
derived from the CAFNl data. Indeed, this latter distribution is similar to
the abstinence distribution obtained from the poorer urban sample (NF2-2).
TABLE 4
Simulated Age-Specific Fertility per loo0 Women*
NF2-2
NF2-1
Age group
10-14
lSl9
20-24
25-29
3&34
35-39
Lactation
0
Abstinence
0
NF2-3
Lactation
Abstinence
Lactation
0
Abstinence
0
4549
5&54
455
1369
1862
1598
1122
458
82
4
376
1010
1258
1105
810
362
65
3
18
912
1990
1902
1618
1145
426
81
6
20
814
1521
1421
1293
897
394
67
7
247
1610
2515
2143
1433
527
105
7
271
1566
2319
1952
1394
474
97
6
Total
6950
4989
8098
6434
8587
8079
29.9
14.0
23.6
6.7
8.4
22
6
18
9
18
7
22
10
22
8
112
M. G. SANTOW
MICROSIMULATION
FIG. 6.
OF YORUBA
FERTILITY
113
40-44 age group the declines are respectively 47 per cent, 30 per cent and 32
per cent.
The variations in the fertility decreases exemplify once again the differential erosion of a traditional practice, for although the fertility of the
urban women has declined, the extent of the decline is not as great as that
of the rural women. The mean completed fertility of each run is significantly lower at the 0.05 level than that of its parent run which does not
TABLE 5
Simulated Age-Specific Fertility of NF2 Sub-samples Showing the Effect of Terminal Abstinence
NF3 Terminal abstinence
Original runs
Age group
NF2-1
NF2-2
NF2-3
NF2-1
NF2-2
NF2-3
10-14
15-19
2&24
25-29
30-34
35-39
4549
5G54
0
376
1010
1258
1105
810
362
65
3
20
814
1521
1421
1293
897
394
67
7
0
251
1496
2188
1923
1274
520
83
7
0
369
973
1259
1139
738
192
25
0
17
798
1486
1438
1241
798
274
16
0
0
228
1456
2186
1875
1146
352
39
0
Total
4989
6434
7742
4695
6068
7282
M. G. SANTOW
114
allow for terminal abstinence, with the falls in mean fertility increasing from
about 0.3 live births for the rural women to 0.4 for the poorer Ibadan
women and to 0.5 for the richer Ibadan women.
8.
CONCLUSIONS
The demonstration
of the impact on fertility of the length of the period
of post partum non-susceptibility
to conception,
and the comparison of
simulated and reported CAFNI fertility data, encouraged an attempt to
simulate the fertility of the three groups of Yoruba women, namely, rural,
poorer ibadan and richer Ibadan. The preliminary simulations were refined
by the inclusion of patterns of terminal abstinence
appropriate
to each
sub-sample.
We may regard the three final simulations as static representations
of the
fertility experience of a population
at different stages in a process of
increasing Westernization
accompanied
by a weakening of the force of
traditional practices. Viewed in this way, the simulations possess dramatic
implications:
the rural fertility of 4.7 live births increases to the poorer
urban fertility of 6.1 iive births, to the richer urban fertility of 7.3 live births.
In centers other than Ibadan one would expect these figures to be slightly
different because of different patterns of marriage, lactation and sexual
abstinence, but the overall conclusion would not change. As women move
from the villages and towns into the city, they can be expected to produce
nearly one and a half more babies than their sisters who stayed at home. As
the financial situation of these urban women improves they can be expected
to produce, on average, slightly more than one additional child, and this
fertility increase would be even greater but for the fact that these women
marry about four years later than their poorer urban counterparts.
In her study of a central Javanese village Hull [9] found a positive
correlation
between the average numbers
of ever born and surviving
children, and income. Women in the 25-44 age group were differentiated by
two patterns of abstinence according to income, with lower-income women
abstaining for about five months longer than upper-income
women. There
was no difference in abstinence levels by income for women in the 15-24
age group, although the periods of stated abstinence were about six months
shorter, at ten months, than those reported by the lower income women in
the 25-34 age group. However, in contrast with this trend, Hull [9] noted
that a later age at marriage of women with higher levels of schooling was
seen to affect cumulative fertility averages of current 2624 year olds.
That urbanization
and economic development lead to a reassessment of
personal goals (or even to the first such assessment)
and hence to a
gFor example,
breastfeeding
Adegbola,
in Lagos
(1) discovered
than Ibadan.
much
shorter
periods
of
MICROSIMULATION
OF YORUBA FERTILITY
115
On the positive side, however, the proportion of women who have never
used family planning fell from 91 per cent of the rural women and 94 per
cent of the poorer urban women, to 30 per cent of the richer urban women.
While only 16 per cent of CAFN 1 women have ever used modem contraception, Caldwell and Caldwell [5] report not only a low dropout rate
amongst users, but also that the level of contraceptive usage has doubled
every four years during the previous twenty. The simulations indicate the
magnitude of the gap left by the breakdown of marital sexual abstinence
which needs to be filled by the use of modem contraceptives
if Yoruba
fertility is not to undergo a dramatic increase.
I am grateful to Professor J. C. Caldwell of the Demography Department,
Australian National University, and to the late Professor F. 0. Okedgi of the
OSee.Robinson [22] and Caldwell [4].
Lucas and Ukaegbu [12] cite a comparable figure of 3 per cent of the rural Ngwa
Ibo.
116
M. G. SANTOW
MICROSIMULATION
OF YORUBA
117
FERTILITY
17 P. 0. Olusanya,
Nigeria:
cultural barriers to family planning
among the Yorubas,
Studies in Family Planning 37: 13-16 (1969).
18 I. 0. Orubuloye,
Family obligations
and fertility in Nigeria: the case of the Yoruba of
western Nigeria,
in The Economic and Social Supports for High Fertili@ (L. T.
Ruzicka, Ed.), Family and Fertility Change: Changing
African Family Companion
Series No. 2, Department
of Demography,
Australian
National University, Canberra,
1977, pp. 203-217.
19 I. 0. Orubuloye and J. C. Caldwell,
a study of mortality
differentials
29:259-272 (1975).
family
size in western
Nigeria,
Journal of Biosociul