• TABLE OF CONTENTS
HIDE
 Table of Contents
 Editorial committee
 Breastfeeding: Patterns, correlates,...
 Family size and sex preferences...
 Sociocultural factors and fertility...
 Healthcare for women in Latin America...














Title: Studies in family planning
ALL VOLUMES CITATION THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00086911/00002
 Material Information
Title: Studies in family planning
Abbreviated Title: Stud. fam. plann.
Physical Description: v. : ill. (part fold.), maps (part col.) ; 28 cm.
Language: English
Creator: Population Council
Publisher: Population Council
Place of Publication: New York
Frequency: quarterly[<1997- >]
monthly[ former july 1963-]
monthly (with combined june/july, aug./sept. issues)[ former ]
bimonthly[ former ]
quarterly
regular
 Subjects
Subject: Birth control -- Periodicals   ( lcsh )
Family Planning Services -- Periodicals   ( mesh )
Gezinsplanning   ( gtt )
Régulation des naissances -- Périodiques   ( rvm )
Genre: periodical   ( marcgt )
 Notes
Additional Physical Form: Also issued online.
Dates or Sequential Designation: v. 1- July 1963-
Numbering Peculiarities: Vol. 1- called also no. 1-<25>
General Note: Supplements accompany some issues.
 Record Information
Bibliographic ID: UF00086911
Volume ID: VID00002
Source Institution: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: oclc - 01651215
lccn - 71001187
issn - 0039-3665

Table of Contents
    Table of Contents
        Page 1
    Editorial committee
        Page 2
    Breastfeeding: Patterns, correlates, and fertility effects
        Page 3
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
    Family size and sex preferences among women in rural Bangladesh
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
    Sociocultural factors and fertility in a rural Nigerian community
        Page 33
        Page 34
        Page 35
    Healthcare for women in Latin America and the Caribbean
        Page 36
        Page 37
        Page 38
Full Text







STUDIES IN FAMILY PLANNING





Volume 12 Number 3
March 1981

Breastfeeding: 79
Patterns, Correlates, and Fertilty Effects
Anrudh K. Jain and John Bongaarts

Family Size and Sex Preferences 100
among Women in Rural Bangladesh
Nilufer R. Ahmed

Sociocultural Factors and Fertility 109
in a Rural Nigerian Community
Adefunke Oyemade and Taiwo A. Ogunmuyiwa

Forum 112
Health Care for Women
in Latin America and the Caribbean,
by Mayra Buvinik and joanne Leslie













STUDIES IN FAMILY PLANNING


Editorial Committee
George F. Brown,
Chairman
Judith Bruce
Ethel P. Churchill
Margaret McEvoy
Susan A. Robbins
S. Bruce Schearer
Irving Sivin
Beverly Winikoff


Editorial Staff
Valeda Slade,
Managing Editor
Robert Heidel,
Project Editor
Renee Santhouse,
Production/Design
Sarah Vure,
Graphic Assistant


Advisory Board


Maria del Carmen
Elu de Lefiero
Consultant, Rural Health and
Family Planning
Mexico City

Juan M. Flavier, M.D.
International Institute
of Rural Reconstruction
Cavite, Philippines

R. H. Gray, M.D.
Department of Population
Dynamics
The Johns Hopkins
University
School of Hygiene
and Public Health
Baltimore, Maryland

Albert I. Hermalin, Ph.D.
Population Studies Center
University of Michigan
Ann Arbor, Michigan

Supom Koetsawang, M.D.
Siriraj Hospital
Mahidol University
Bangkok, Thailand

Guillermo Lopez-Escobar, M.D.
Corporacidn Centro
Regional
de Poblaci6n
Bogota, Colombia


Firman Lubis, M.D.
University of Indonesia
Jakarta, Indonesia
Olikoye Ransome-Kuti, M.D.
Department of Pediatrics
College of Medicine
University of Lagos
Lagos, Nigeria
Susan C.M. Scrimshaw, Ph.D.
Division of Population,
Family and International
Health
School of Public Health
University of California
at Los Angeles
Belgin Tekce, Ph.D.
Social Research Center
The American University
in Cairo
Cairo, Egypt
Miriam K. Were, M.D.
Department of Community
Health
Faculty of Medicine
University of Nairobi
Nairobi, Kenya
Huda Zurayk, Ph.D.
Department of Epidemiology
and Biostatistics
American University
of Beirut
Beirut, Lebanon


Studies in Family Planning
Studies is an international journal devoted to articles and
commentary on all aspects of fertility control and family
planning, with particular emphasis on activities in the de-
veloping countries. Areas of current interest include
advances in contraceptive technology, efficacy, and safety;
the interface of women's roles and status with family plan-
ning; the consumer's perspective and contraceptive
choice; and the health, social, political, and economic im-
pacts of fertility regulation policies and programs.

Submission of Papers
Manuscripts are invited and should be submitted in tripli-
cate to the Managing Editor. An abstract of 100 words or
less, authors' titles and affiliations, and acknowledgments
should be included with the manuscript. All manuscripts
are refereed by professionals in the field.

Subscriptions
Studies is available upon request to all qualified individual.
and institutions. Subscription and change of address re-
quests should be sent to the Circulation Supervisor.
Subscribers should include their current mailing label with
all correspondence.

Binders
Binders for the volume year are available at no charge to
subscribers in developing countries (one per person), and
at a prepaid charge of US$3.50 each to subscribers in
developed countries.

Indexes
Indexes to Studies are published annually in a separate
section distributed with the December issue. Cumulative
indexes are available for Volumes 1-3 (July 1963-
December 1972), Volumes 4-6 (1973-1975), and Vol-
umes 7-9 (1976-1978).

Spanish Edition
Spanish translations of selected articles from Studies, as
well as original articles, are published in Estudlos de
Poblacl6n, issued by the Asociaci6n Colombiana para el
Estudio de la Poblaci6n (ACEP), Carrera 23, No. 39-82,
Bogota D.E., 1, Colombia.

Studies in Family Planning is published monthly with
combined issues in june/July and August/September by
the Population Council, One Dag Hammarskjold Plaza,
New York, NY 10017, USA. Copyright 1981 by the
Population Council, Inc. All rights reserved. Opinions ex-
pressed in Studies do not necessarily reflect the views of
the Population Council. Application to mail at second-clas
postage rates pending at New York, NY, and additional
mailing offices. Postmaster: Send address changes to
Studies In Family Planning, One Dag Hammarskjold
Plaza, New York, NY 10017.


ISSN 0039-3665









Breastfeeding:

Patterns, Correlates, and Fertility Effects

Anrudh K. Jain and John Bongaarts


Fertility is directly influenced by a set of so-
ciobiological factors. These factors are often called
intermediate fertility variables' because they are in-
fluenced in turn by various economic, social, cultural,
and environmental variables (which are the indirect
or background determinants of fertility). A recent
study of the fertility effects of the intermediate fertil-
ity variables has demonstrated that nearly all variance
in the fertility levels of populations is due to differ-
ences in just four factors: (1) the proportions married
among females, (2) the prevalence of contraceptive
use, (3) the incidence of induced abortion, and (4) the
fertility inhibiting effect of breastfeeding.2
The importance of breastfeeding in regulating in-
dividual fertility behavior has been a matter of inter-
est for many years. The lack of uniform data for more
than one country has, so far, limited the scope of
cross-cultural analysis of breastfeeding and its deter-
minants. The data generated through the World Fer-
tility Survey provide us with a unique opportunity to
understand the behavior of women with respect to
breastfeeding and its influence on fertility on a cross-
cultural comparative basis. This paper will address
the following questions:
1 What is the prevalence and duration of breast-
feeding?
2 Is the preference for male children supported by
differential breastfeeding patterns for male and
female children?
3 How does the duration of breastfeeding vary
among different subgroups classified by age,
parity, women's education, residence, and so
on?

Anrudh K. Jain, Ph.D.,.is Senior Associate, International
Programs, the Population Council. John Bongaarts, Ph.D.,
is Associate, Center for Policy Studies, the Population
'Council.


4 What are the key determinants of breastfeeding?
5 Do women use breastfeeding deliberately to
space or limit the number of children?
6 What is the effect of breastfeeding on fertility?


The Data

The data for this study are taken from the core ques-
tionnaires of the World Fertility Surveys conducted
around 1976 in eight countries: Bangladesh, Indo-
nesia, Sri Lanka, Jordan, Peru, Guyana, Colombia,
and Panama. Data tapes for these countries were
made available to the authors, and SPSS (Statistical
Package for Social Sciences) and other special pro-
grams written by Robert Sendek were used in this
analysis. The limited data on breastfeeding included
in the First Country Reports are not comparable to
the information presented here and, therefore, are
not included in this paper.
Information on the prevalence and duration of
breastfeeding was collected for the two live births im-
mediately preceding the interview. Among currently
pregnant women, breastfeeding data for the next-to-
last birth were not available. Only data for the last-
but-one live birth are used here for studying the
determinants of breastfeeding and its influence on
fertility. This is done because the interview truncated
the women's reproductive history, and the informa-
tion about breastfeeding in the open birth interval
was not complete. We have, however, used this in-
formation to estimate the mean and median duration
of breastfeeding in the open birth interval.3
The unavailability of data for some women made
it necessary to limit analysis to women who were
married at the time of the interview, had two or more
live births, had reported the duration of breastfeed-
ing, were not pregnant at interview, and had their
last-but-one live birth between three and 15 years


Studies in Family Planning Volume 12 Number 3 March 1981 79








preceding the date of interview. The last restriction
was used to minimize the effects of truncation and
memory biases on the reported duration of breast-
feeding and the length of the birth interval. (It would
have been preferable to further restrict this period to
perhaps 3-8 years, but that would have further re-
duced the number of women included in the analy-
sis.) With the current restrictions, the analysis based
on the information about the last-but-one live birth
refers to 28 percent of women included in the original
surveys in Guyana and Colombia; 42 percent in
Bangladesh, Indonesia, Jordan, and Panama; and 49
percent in Peru and Sri Lanka. This limited sample
will be referred to as all women in the rest of this pa-
per.
The effect of breastfeeding on fertility is meas-
ured by using the last closed birth interval as the
proxy for the fertility level. The last closed birth inter-
val is defined as the period in months between the
last-but-one live birth and the last live birth preced-


ing the interview. Since the World Fertility Surveys
did not collect information about the date of resump-
tion of menstruation, this study cannot analyze the
mechanisms through which breastfeeding affects the
length of the birth interval.
The results presented below are necessarily in-
fluenced by the biases associated with the retrospec-
tive nature of the data collection. The magnitude of
the bias in reporting the duration of breastfeeding is
likely to differ from country to country and from one
subgroup to another within the same country. There
is a strong tendency to report the duration of breast-
feeding in multiples of six months in all countries. In
the absence of any such tendency, one-sixth or about
16 percent of the women are likely to report the dura-
tion of breastfeeding in multiples of six months. In
comparison, this proportion is about 34 percent in
Panama and Colombia; 45 percent in Guyana, Peru,
and Jordan; about 60 percent in Sri Lanka and Indo-
nesia; and about 86 percent in Bangladesh. There


FIGURE 1 Percent distribution of women by duration of breastfeeding, selected countries


80 Studies in Family Planning









may be cultural preferences or norms to breastfeed a
child for 12 or 24 months. In that case, the difference
between the observed and the expected percent of
women who reported the duration of breastfeeding
in multiples of six months cannot be attributed en-
tirely to the digital preferences. Allowing for some
cultural preferences, the bias in reporting the months
of breastfeeding does not seem to be serious in Pan-
ama and Colombia, whereas it is quite considerable in
Bangladesh, Indonesia, and Sri Lanka. The differ-
ences in the percent distribution of women by
months of breastfeeding reflect the variations in the
reporting biases (see Figure 1). Given the magnitude
of the bias, we did not estimate the duration of
breastfeeding by year of birth of the child, and did
not estimate the time trends in prevalence or duration
of breastfeeding.
The eight countries included in this study are not
a random selection of countries and, therefore, the


result of this study cannot necessarily be generalized.
Some of the findings may need to be modified as sim-
ilar studies for other countries become available. No
African country could be included, but the eight
countries are quite heterogeneous in many respects-
for example, geographic area, religion, culture, fertil-
ity, and level of development.
In Table 1 we show the percentage distributions
of all women by selected social and demographic fac-
tors. (Table 2 shows similar information for nonusers
of contraception.) The composition of women differs
markedly from country to country. For example, 77
percent of all women in Bangladesh had no educa-
tion, compared with only 4 percent in Guyana; 78
percent of all women in Bangladesh lived in rural
areas, compared with 31 percent in Jordan; and about
85 percent of the women in Bangladesh and Jordan
were classified as not working since marriage, com-
pared with 35 percent in Indonesia and Peru. The dif-


TABLE 1 Percent distribution of all women, by selected demographic and social characteristics
Demographic and social Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
characteristics (N=2660) (N=4064) (N=3399) (N=1521) (N=2711) (N=1276) (N=1537) (N=1556)
Age of wife
15-19 25.6 15.4 8.4 7.1 9.2 10.2 13.0 10.4
20-24 27.2 25.1 23.5 18.1 24.0 27.4 26.2 32.6
25-29 19.9 25.2 29.5 24.5 23.6 27.4 26.5 28.5
30-34 16.9 21.9 25.1 28.5 24.0 19.5 19.0 17.7
35-39 9.0 10.4 12.2 18.5 15.0 12.6 13.0 9.2
40+ 1.4 2.0 1.4 3.3 4.1 2.9 2.3 1.5
Parity
2-3 29.2 32.9 31.4 15.8 28.3 27.3 32.8 37.0
4-6 38.3 40.8 39.4 28.7 35.3 39.5 33.3 38.1
7+ 32.5 26.3 29.2 55.6 36.4 33.2 33.9 24.9
Wife's education
None 76.8 60.2 23.0 62.5 33.6 4.2 17.0 8.6
Primary 19.2 30.2 40.5 27.6 48.4 76.0 68.6 57.0
Secondary+ 4.0 9.7 36.5 9.9 18.0 19.8 14.5 34.4
Residence
Rural 77.8 68.7 74.4 31.0 36.5 67.9 35.4 47.9
Urban 22.2 31.3 25.6 69.0 63.5 32.1 64.6 52.1
Work place of wife (since marriage)
Family farm .5 23.6 11.6 4.5 22.4 7.0 2.6 1.8
Other farm .9 13.8 3.0 1.2 3.2 2.9 2.0 .4
At home 3.9 10.2 3.6 3.5 17.9 6.7 11.8 5.8
Away from home 7.4 17.8 19.8 5.3 21.4 29.7 22.6 40.8
Did not work 87.4 34.7 62.0 85.4 35.2 53.7 61.0 51.1
Husband's occupation
Professional and
clerical 9.9 12.1 11.6 18.1 14.9 15.3 9.8 15.6
Sales and service 14.6 19.0 15.6 21.0 14.4 4.8 16.4 17.7
Skilled and manual 22.3 18.9 18.9 31.0 18.7 33.2 33.5 34.2
Farmer and
agricultural 50.8 49.1 41.4 10.8 42.3 36.8 36.7 28.9
Other 2.4 .9 12.5 19.2 9.7 10.0 3.7 3.6


Volume 12 Number 3 March 1981 81








ferences between countries in the composition of
women with respect to social factors are more pro-
nounced than those relating to demographic factors.
The effects of these differences on the duration of
breastfeeding will be studied in a later section.



Prevalence
and Duration
of Breastfeeding

Table 3 shows various indices measuring the preva-
lence and duration of breastfeeding. These indices
are shown separately for the last live birth (open birth
interval) and for the last-but-one live birth (the last
dosed birth interval). Countries are arranged in de-
creasing order according to the average duration of
breastfeeding for the closed birth interval.


In all eight countries, a large majority of women
breastfed their last as well as their last-but-one child.
The proportion of women who did not breastfeed
their last child ranged from 2 percent in Bangladesh
to 18 percent in Panama.
In all countries, the average duration of breast-
feeding estimated from breastfeeding status in the
open birth interval is higher than the corresponding
average for the closed birth interval. The difference
between the two averages is unlikely to be caused by
an increase in the duration of breastfeeding over
time. It is more likely to reflect the biases in reporting
the duration of breastfeeding in the closed birth inter-
val, and to some extent an improvement in infant
mortality and perhaps differences in samples.
There is a great deal of variation between coun-
tries with respect to the duration of breastfeeding.
Women in Bangladesh breastfed their children for
about 29 months, compared with about nine months


TABLE 2 Percent distribution of women who did not use contraception during the last closed birth interval,
by selected demographic and social characteristics
Demographic and social Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
characteristics (N=2483) (N=3241) (N=2772) (N=993) (N=1876) (N=1041) (N=994) (N=993)
Age of wife
15-19 25.5 16.5 8.6 7.5 9.1 10.2 12.7 11.2
20-24 27.2 24.2 23.1 14.5 19.9 26.2 22.2 26.8
25-29 19.5 24.0 28.8 21.4 21.6 26.2 24.1 27.2
30-34 17.2 22.4 25.3 30.8 25.9 20.4 21.6 22.1
35-39 9.2 10.8 12.6 22.0 18.5 13.6 16.2 10.8
40+ 1.4 2.1 1.6 3.8 5.1 3.4 3.1 1.9
Parity
2-3 29.0 32.4 29.4 14.2 24.1 25.0 26.8 31.4
4-6 38.1 39.9 39.4 26.3 33.8 38.7 32.9 38.2
7+ 32.9 27.7 31.2 59.5 42.1 36.3 40.3 30.4
Wife's education
None 79.2 64.3 26.0 77.2 42.3 5.0 23.4 11.3
Primary 18.4 29.4 41.7 19.8 46.3 78.4 69.4 64.6
Secondary+ 2.4 6.3 32.3 3.0 11.4 16.6 7.2 24.1
Residence
Rural 79.8 70.8 75.8 42.3 45.5 71.8 44.1 57.4
Urban 20.2 29.2 24.2 57.7 54.5 28.2 55.9 42.6
Work place of wife (since marriage)
Family farm .4 23.3 11.9 6.6 28.3 8.0 3.4 2.4
Other farm .9 14.8 3.5 1.9 3.8 3.4 2.7 .4
At home 4.0 10.0 3.1 2.7 17.2 6.5 10.3 6.3
Away from home 7.2 17.0 19.9 2.0 17.2 26.9 20.4 32.8
Did not work 87.5 34.9 61.6 86.8 33.4 55.2 63.1 58.1
Husband's occupation
Professional and clerical 8.5 9.4 10.1 11.6 10.2 12.5 5.2 11.1
Sales and service 13.8 19.1 15.2 21.0 11.6 4.5 14.0 16.9
Skilled and manual 22.6 19.7 17.8 29.0 16.4 33.2 30.8 30.8
Farmer and agricultural 52.7 50.8 44.0 15.3 52.5 38.9 45.9 37.3
Other 2.3 1.0 12.9 23.1 9.3 10.9 4.1 3.9


82 Studies in Family Planning








in Panama. The longer duration of breastfeeding in
Bangladesh is also reflected by the fact that 61 percent
of these women reported they were breastfeeding
their last child at interview, compared with only 15
percent in Panama.



Determinants of
Breastfeeding

Sex and Survival Status
of Last-But-One Live Birth
In Table 4 we show the average duration of breast-
feeding by (1) sex of the last-but-one child, (2)
whether or not s/he survived until the interview, and
(3) use of contraception in the last closed birth inter-
val.
It is believed that in some developing countries,
female children are neglected because of a strong
preference for male children. If this neglect exists,
one would expect it to result in shorter breastfeeding
and higher infant and child mortality among females.
Results presented in Table 4 do not support this hy-
pothesis. The average duration of breastfeeding for
male children is about the same as for female children
in all eight countries. There was also no sex differen-
tial in the average duration of breastfeeding for chil-
dren who survived up to the time of interview


(results not shown here). These children were at least
three years old at the time of interview. Any sex dif-
ferentials in child care practices beyond breastfeeding
are of course not reflected in these results.
Death of a child curtails the period of breastfeed-
ing. Results presented in Table 4 confirm this hypoth-
esis. The average duration of breastfeeding for those
who died in infancy was much shorter than for those
who died at a later age or those who were alive at the
time of interview. In all countries except Colombia,
the reported duration of breastfeeding is not consis-
tent with the reported age of death for those who
died at age 0 months. This inconsistency reflects re-
porting error and is especially serious in Bangladesh
and Jordan. For children who were alive at interview,
the average duration of breastfeeding is slightly
higher than the corresponding averages for all chil-
dren.

Use of Contraception
The reported use of contraception during the last
closed birth interval varies from 6 percent in Bangla-
desh to 40 percent in Panama. Among women who
did not use contraception, the average duration of
breastfeeding varies from 10 months in Panama, Co-
lombia, and Guyana to 24 months in Bangladesh. The
average duration of breastfeeding among those who
used contraception is generally lower than among
those who did not use contraception during the last


TABLE 3 Selected statistics on breastfeeding in open and last closed birth interval
Statistics Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
Open birth intervals
Percent who did
not breastfeed 2 4 5 8 10 N.A. 9 18
Mean 28.8 23.2 22.0 13.1 14.2 N.A. 10.0 9.0
Median 29 21 21 14 14 N.A. 11 10
Percent currently
breastfeeding 61 41 36 39 32 N.A. 19 15
Nb 3230 5245 4411 2311 3742 1912 2035 2071
Closed birth inter-
val
Percent who did
not breastfeed 4 4 6 8 10 N.A. 10 16
Mean 23.6 19.0 15.7 12.5 11.7 10.0 8.6 8.3
S.D. 11.7 10.8 11.2 8.4 8.9 8.6 8.1 8.6
Nc 2660 4064 3399 1521 2711 1276 1537 1556
NOTE: Women who did not breastfeed are assigned the value zero in calculating mean and standard deviation of breastfeeding.
aSee note 3 for estimation of mean and median, which are derived from breastfeeding status at interview of all births during four years
preceding interview.
bNumber of currently married women with two or more live births; excludes unknown BF.
'Number of currently married women with two or more live births; excludes unknown BF; excludes if CBI + OBI -36 or 180 months, or if
Currently pregnant.


Volume 12 Number 3 March 1981 83











TABLE 4 Average duration of breastfeeding by sex and survival status of the child and use of contraception
during the last closed birth interval

Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama

Average duration of breastfeeding (months)

Total average 23.6 19.0 15.7 12.5 11.7 10.0 8.6 8.3
Sex of child
Male 23.9 18.7 15.6 12.8 11.7 10.0 8.7 8.7
Female 23.3 19.3 15.8 12.2 11.7 10.0 8.5 7.7
Survival status
Dead: Age-
0 months 14.9 2.2 2.9 5.8 1.3 1.7 0.6 2.2
5 1 year 14.0 4.1 3.5 3 3.3 3.6 2.1 3.2
1 year 19.2 15.6 14.5 10.2 10.6
Alive: 24.5 20.4 16.4 13.1 12.3 10.3 9.0 8.5
Use of contraception
No method 23.6 19.2 15.7 14.1 12.8 10.7 9.5 9.8
Inefficient method 25.9 20.3 17.3 9.9 10.8 6.3 7.7 8.6
Efficient method 21.0 17.0 13.5 9.5 6.5 7.0 6.3 5.1
Percent distribution of women

Sex of child
Male 51 51 51 55 50 54 51 53
Female 49 49 49 45 50 46 49 47
Survival status
Dead: Age-
0 months 7 4 3 4 4 2 3 3
s 1 year 2 4 2 3 4 3 3 1
1 year 6 3 4 4 1
Alive: 91 86 92 93 88 95 90 95
Use of Contraception
No method 94 80 82 65 69 82 65 60
Inefficient method 3 9 12 11 21 2 15 11
Efficient method 3 11 6 24 10 16 20 29

*Included in Alive category.



TABLE 5 Summary of multiple regression analysis using duration of breastfeeding as the dependent
variable for all women and for those who did not use contraception during the last closed birth interval

Demographic
characteristics Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
I. All women

Correlation
coefficient
Age -.026 .017 -.005 .121 .143 .050 .171 .190
Parity -.101 -.031 .039 .104 .153 .106 .155 .275
Partial
regression coefficient
Intercept 22.10 17.15 17.00 8.05 7.41 9.31 3.47 3.91
Age .299* .151* -.112* .137* .090* -.063 .155* .0
Parity -1.09 -.426* .341* .070 .308* .423* .165 .880*
R2 .020 .005 .003 .015 .025 .012 .031 .076

II. Women who did not use contraception


Correlation
coefficient
Age
Parity


-.027
-.104


.007
-.041


84 Studies in Family Planning











TABLE 5 (continued)

Demographic
characteristics Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
II. Women who did not use contraception
Partial
regression coefficient
Intercept 22.14 17.72 16.99 10.03 8.47 10.20 4.29 5.54
Age .304* .140* -.106* .179* .125* -.046 .178* .019
Parity -1.115* -.446* .291* -.176 .109 .293* .042 .688*
R2 .021 .005 .003 .009 .016 .006 .026 .054

*Regression coefficient is greater than twice its standard error.




TABLE 6 Effect of wife's age on duration of breastfeeding, unadjusted and adjusted through Multiple
Classification Analysis, for the effects of wife's education, place of residence, work place, and husband's
occupation, for all currently married women and for those who did not use contraception during the last
closed birth interval

Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama

I. All women

Grand mean, months
of breastfeeding 23.6 19.0 15.7 12.5 11.7 10.0 8.6 8.3
Wife's age A. Deviations from grand mean (unadjusted)
15-19 .6 .5 .3 -1.3 .8 .5 .9 .8
20-24 .3 .6 .2 -1.6 -1.4 -1.4 -1.7 -1.6
25-29 .2 .1 .7 .8 -1.0 .5 .2 .5
30-34 .4 .2 .7 1.2 .8 .1 1.1 2.2
35-39 .6 .7 .5 1.4 2.8 1.2 2.6 3.3
40+ -2.8 .2 1.1 .2 1.0 .4 3.0 4.6*
B. Deviations from grand mean (adjusted)
15-19 .7 .0 .1 .8 .2 1.4 .6 .5
20-24 .2 .4 .4 .5 .1 -1.0 -1.1 -1.0
25-29 .2 .3 .3 .3 .1 .3 .0 .1
30-34 .3 .1 .5 .7 .1 .3 .6 1.3
35-39 .8 .2 .9 .5 1.0 .7 1.6 1.6
40+ -3.4 -1.4 .7 -1.6 -2.0 .1 2.2 1.7*
II. Women who did not use contraception

Grand mean, months
of breastfeeding 23.7 19.2 15.6 14.1 12.8 10.6 9.6 9.8
Wife's age A. Deviations from grand mean (unadjusted)
15-19 .7 .6 .2 -1.5 -1.2 1.1 -1.3 -1.2
20-24 .4 .7 .4 -1.1 -1.2 -1.7 -1.6 -1.3
25-29 .5 .1 .7 .7 .9 .8 .5 .9
30-34 .7 .0 .7 .7 .5 .1 1.2 1.4
35-39 .6 .5 .6 1.1 2.2 .9 1.7 3.4
40+ -2.8 .4 1.0 .6 .3 0 2.9 3.2*
B. Deviations from grand mean (adjusted)
15-19 .7 .3 .2 -1.1 .6 2.0 -1.2 .7
20-24 .3 .5 .6 .6 .1 -1.3 -1.1 -1.0
25-29 .4 .2 .4 .7 .1 .6 .3 .5
30-34 .5 .1 .5 .6 .1 .4 .9 1.0
35-39 .7 .3 .9 .8 1.1 .5 1.1 2.2
40+ -3.2 -1.0 .7 -1.2 -1.6 .2 2.3 1.6*

*Less than 25 cases.


Volume 12 Number 3 March 1981 85










TABLE 7 Effects of parity on duration of breastfeeding, unadjusted and adjusted through Multiple Classifi-
cation Analysis, for the effects of wife's age, education, place of residence, work place, and husband's
occupation, for all women and for those who did not use contraception during the last closed birth interval
Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
I. All women
Grand mean, months
of breastfeeding 23.6 19.0 15.7 12.5 11.7 10.0 8.6 8.3
Parity A. Deviations from grand mean (unadjusted)
2-3 1.1 .0 -1.3 -2.8 -2.7 -1.6 -1.8 -2.8
4-6 .2 .5 .6 .5 .7 .2 .1 .4
7+ -1.2 .7 .5 1.1 1.5 1.1 1.6 3.5
B. Deviations from grand mean (adjusted)
2-3 2.4 .7 .5 -1.6 .4 .8 .3 -1.4
4-6 .3 .4 .6 .1 .9 .1 .2 .4
7+ -2.5 -1.6 .2 .5 .6 .6 .1 1.4
II. Women who did not use contraception
Grand mean, months
of breastfeeding 23.7 19.2 15.6 14.1 12.8 10.6 9.6 9.8
Parity A. Deviations from grand mean (unadjusted)
2-3 1.2 .2 -1.4 -2.6 -2.3 -1.1 -1.6 -2.7
4-6 .1 .4 .9 .1 .5 .1 .1 0
7+ -1.2 .8 .2 .6 .9 .7 1.2 2.8
B. Deviations from grand mean (adjusted)
2-3 2.4 .9 .8 -1.6 .2 .7 .2 -1.6
4-6 .2 .3 .9 .1 .8 0 .3 .1
7+ -2.4 -1.5 .4 .3 .7 .4 .1 1.5


closed birth interval. This relationship between use of
contraception and duration of breastfeeding will be
explored in more detail later.

Age and Parity

In earlier studies, mother's age has been found to
have a positive influence on the duration of breast-
feeding.4 If breastfeeding is used deliberately to limit
family size, its duration should be affected by the
number of children already born.5 For Taiwanese
women, however, Jain et al.6 found that in a multiple
regression analysis, women's parity did not have any
significant effect on the duration of breastfeeding af-
ter controlling for the effects of such factors as
women's age, education, and place of residence.
For the eight countries included in this analysis,
the effect of age and parity is not important (age is
measured at the beginning of the closed birth inter-
val). There is no consistent pattern; that is, the direc-
tion as well as the magnitude of these effects depends
upon country of residence. Among those who did not
use contraception, mother's age and parity explain
less than one percent of the variation in the duration


of breastfeeding in Indonesia, Sri Lanka, Jordan, and
Guyana and about 2 percent in Bangladesh, Peru,
and Colombia. Only in Panama is the percent varia-
tion explained by age and parity slightly more than 5
percent (see Table 5). The partial regression coeffi-
cients indicate that the net effect of mother's age on
the duration of breastfeeding, among those who did
not use contraception, is not statistically significant
for Guyana and Panama. In Sri Lanka, the net effect
of age is negative, whereas in the remaining five
countries the net effect of age is positive. In all coun-
tries, however, the net effect of age on the duration of
breastfeeding is small. For example, in Bangladesh,
about three years' increase in mother's age adds
about one month to the duration of breastfeeding; in
Peru, about eight years' increase in mother's age adds
about one month to the duration of breastfeeding.
The net effect of parity, on the other hand, is not sta-
tistically significant in Jordan, Peru, and Colombia; it
is negative in Bangladesh and Indonesia, and it is
positive in the remaining three countries-Sri Lanka,
Guyana, and Panama.
The results of Multiple Classification Analysis
are shown in Tables 6 and 7. The category means are


86 Studies in Family Planning










TABLE 8 Effects of wife's education and place of residence on duration of breastfeeding, unadjusted and
adjusted through Multiple Classification Analysis, for place of residence or education and for wife's age,
parity, work place, and husband's occupation, for all women and for those who did not use contraception
during the last closed birth interval
Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
I. All women

Grand mean, months
of breastfeeding 23.6 19.0 15.7 12.5 11.7 10.0 8.6 8.3
A. Deviations from grand mean (unadjusted)
Education of wife
None .4 1.5 1.6 1.7 3.8 4.3 3.1 6.0
Primary .2 .4 1.5 -1.7 .4 .4 .1 1.5
Secondary+ -6.7 -8.3 -2.7 -8.0 -6.1 -2.6 -3.9 -4.0
Residence
Rural .8 2.4 1.0 2.9 3.8 1.2 2.1 2.7
Urban -2.7 -5.2 -2.9 -1.3 -2.2 -2.6 -1.1 -2.5
B. Deviations from grand mean (adjusted)
Education of wife
None .2 .8 .9 .9 2.4 3.5 2.1 3.2
Primary 0 .1 1.0 .8 .2 .2 0 .7
Secondary+ -4.6 -4.8 -1.7 -3.8 -4.0 -1.4 -2.5 -2.0
Residence
Rural .6 1.3 .6 1.4 1.5 .8 .3 .5
Urban -2.0 -2.8 -1.7 .6 .9 -1.8 .2 .5
II. Women who did not use contraception

Grand mean, months
of breastfeeding 23.7 19.2 15.6 14.1 12.8 10.6 9.6 9.8
A. Deviations from grand mean (unadjusted)
Education of wife
None .3 1.0 1.4 .7 2.8 3.9 2.1 4.3
Primary .3 .5 1.2 -1.5 -1.0 .2 .3 .8
Secondary+ -6.3 -8.1 -2.7 -7.0 -6.5 -2.1 -4.3 -4.2
Residence
Rural .7 1.9 1.0 1.7 3.0 .9 1.6 2.1
Urban -2.6 -4.5 -3.0 -1.2 -2.5 -2.4 -1.2 -2.8
B. Deviations from grand mean (adjusted)
Education of wife
None .1 .6 .8 .2 1.7 3.2 1.4 2.2
Primary .0 .3 .7 .3 .6 .1 .2 .3
Secondary+ -4.0 -5.1 -1.6 -4.5 -3.8 -1.3 -2.8 -1.8
Residence
Rural .5 1.0 .6 1.0 1.4 .7 .1 .7
Urban -2.1 -2.5 -1.9 .7 -1.2 -1.8 .1 .9
= less than 25 cases; 0 = less than .05.


expressed as deviations from the grand mean. The
unadjusted deviations indicate the gross effects, and
the adjusted deviations indicate the net effects. The
magnitudes of adjusted deviations are again very
small. For example, the difference in adjusted devia-
tions between any two consecutive categories of age
rarely exceeds two months. The maximum difference
between any two categories is about four months.
The results presented so far indicate that the duration


of breastfeeding is virtually independent of parity. It
is therefore likely that breastfeeding in these eight
countries is not used deliberately to limit family size.


Social Factors

Four social factors included in this study are:
mother's education, place of residence, her work
place since marriage, and her husband's occupation.


Volume 12 Number 3 March 1981 87









TABLE 9 Average duration of breastfeeding (months) by wife's education and place of residence for all
women
Place of
residence Education Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
Rural Total 24.4 21.4 16.7 15.5 15.6 11.1 10.6 11.0
None 24.5 21.8 17.6 15.9 16.8 14.4 12.1 15.1
Primary 24.2 21.0 18.1 14.9 13.8 11.2 10.1 11.0
Secondary+ 20.7* 15.8 13.8 10.7 9.3 8.9 5.2
Urban Total 20.9 13.8 12.8 11.2 9.5 7.4 7.5 5.8
None 21.9 16.3 15.7 13.2 13.2 15.8* 11.2 10.3*
Primary 21.2 13.8 13.7 10.4 10.3 7.9 7.8 7.9
Secondary+ 15.8 9.5 11.7 6.4 5.5 6.0 4.4 4.1
Total Total 23.6 19.0 15.7 12.5 11.7 10.0 8.6 8.3
None 24.0 20.5 17.3 14.2 15.5 14.3 11.7 14.3
Primary 23.4 18.6 17.2 10.8 11.3 10.4 8.7 9.8
Secondary+ 16.9 10.7 13.0 6.5 5.6 7.4 4.7 4.3
*Less than 25 cases.


The results are shown in Tables 8-11. Except in Table
9, the net effect of any one of the four social factors is
the effect of that social factor (for example, education)
on the duration of breastfeeding after adjusting for
the effect of the remaining three social factors (resi-
dence, work place, and husband's occupation) and
two demographic factors (age and parity).
Education and residence: In all eight countries, edu-
cation and urban residence are associated with a
shorter duration of breastfeeding. A similar result
was obtained for Taiwanese women.7 The adjusted
deviations in Table 8 show that the difference in the
average duration of breastfeeding between women
with no education and those with at least seven years
of schooling (secondary +) is from four to six months
for all countries except Sri Lanka, where the differ-
ence is about two months. The difference between
rural and urban areas, on the other hand, is on the
order of two to four months, except in Colombia,
where the difference is less than one-half month.
In Table 9 we show the average duration of
breastfeeding by mother's education separately for
rural and urban areas. It can be seen that both place
of residence and education have independent nega-
tive effects on the duration of breastfeeding. The
average duration of breastfeeding is longest for
women who have no education and live in rural
areas, and it is shortest for those who live in urban
areas and have at least seven years of schooling. The
remaining women fall in between these two ex-
tremes.
Women's work place since marriage: The variable


measuring women's work place since marriage has
been classified by WFS into five categories. This
standard classification combines two dimensions of
work place: (1) farm versus nonfarm and (2) home
versus outside the home. However, there must have
been considerable variation across countries in the
definition or interpretation of "work," because the
percent of women classified in the "did not work"
category varies in an implausible way across coun-
tries: about 35 percent in Peru and Indonesia; be-
tween 51 and 62 percent in Guyana, Panama,
Colombia, and Sri Lanka; and about 87 percent in Jor-
dan and Bangladesh (see Table 1). If modernization
implies a decrease in the duration of breastfeeding (as
indicated by the effects of mother's education and
place of residence), then one would expect that
women who worked away from home in nonfarm
settings would have the shortest duration of breast-
feeding, and those who worked on the family farm
would have the longest duration. However, the net
effects, shown in Table 10, indicate that the inde-
pendent effect of work status or place of work on the
duration of breastfeeding is very small. We cannot
ascertain whether this lack of effect is real or due to
some problems in the definition or the interpretation
of "work" and "place of work."
Husband's occupation: We have regrouped 8-10
standard WFS categories of husband's occupation
into five categories as shown in Table 11. Among
these, the first four categories are of particular inter-
est. These four categories in general can be arranged
in increasing order according to the observed average



88 Studies in Family Planning










TABLE 10 Effects of wife's work place on duration of breastfeeding, unadjusted and adjusted for wife's age,
parity, education, place of residence, and husband's occupation, for all women and for those who did not
use contraception during the last closed birth interval
Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
I. All women
Grand mean, months
of breastfeeding 23.6 19.0 15.7 12.5 11.7 10.0 8.6 8.3
Wife's work place
(since marriage) A. Deviations from grand mean (unadjusted)
Family farm 6.0* 3.8 3.4 2.4 3.4 2.1 2.1 9.2
Other farm .1* 2.3 .2 3.5* 3.6 1.6 3.0 3.6*
At home 2.3 -1.0 2.6 0 .5 .3 -1.4 .8
Away from home -1.3 -2.1 -1.1 -4.6 -2.7 -1.5 -1.0 -2.5
Did not work 0 -2.2 .4 .1 .6 .5 .4 1.6
B. Deviations from grand mean (adjusted)
Family farm 4.8* .9 1.6 .7 .1 .6 .6 4.2
Other farm -1.3* .2 -2.0 .2* 1.1 .3 .9 -1.3
At home 1.9 .1 3.0 .5 .2 .2 .8 .5
Away from home -1.0 .3 -1.4 .1 0 .4 .1 .8
Did not work 0 .4 .1 0 .1 .1 .2 .5
II. Women who did not use contraception
Grand mean, months
of breastfeeding 23.7 19.2 15.6 14.1 12.8 10.6 9.6 9.8
Wife's work place
(since marriage) A. Deviations from grand mean (unadjusted)
Family farm 5.4* 3.2 3.2 .6 2.5 1.3 1.4 8.3*
Other farm .1* 1.7 .2 1.9* 2.1 1.3 2.5 -1.0*
At home 2.1 -1.4 1.2 .6 .6 .7 -1.6 1.1
Away from home -1.3 -1.5 -1.4 -2.6* -3.2 -1.0 -1.0 -2.7
Did not work 0 -1.7 .2 0 .4 .3 .4 1.1
B. Deviations from grand mean (adjusted)
Family farm 4.2* .8 1.6 -1.2 0 .2 .5 4.6*
Other farm -1.2* .4 -1.8 .3* .6 .3 1.0 -5.4
At home 1.8 .3 1.6 .2 0 0 .9 .7
Away from home -1.1 .1 -1.9 -1.5* .4 .1 .1 -1.0
Did not work 0 .3 .3 .1 .1 0 .2 .3
= less than 25 cases; 0 = less than .05.


duration of breastfeeding: (1) professional and cleri-
cal, (2) sales and services, (3) skilled and manual, and
(4) farmers and agricultural. The unadjusted devia-
tions show that the observed differences between the
average duration of breastfeeding for women whose
husband's occupation fell in the first and the fourth
category is about 4-7 months. A large proportion of
the differences is accounted for by the association be-
tween husband's occupation and wife's characteris-
tics, such as her education. (The adjusted deviations
are much smaller than the unadjusted deviation.)
Nevertheless, husband's occupation seems to have a
consistent independent effect on the breastfeeding
behavior of the women in these countries.


Multiple Regression
Analysis

The effects of seven demographic and social factors
on the duration of breastfeeding are summarized in
Table 12 by using multiple regression analysis. Two
multiple regression equations are shown: one for all
women and another only for those who did not use
contraception during the closed birth interval.
Mother's age is measured in single years and parity in
single number of live births. The remaining factors
are included as dummy variables. (Infant death is as-
signed a value of one if the child died before reaching
age one year; otherwise it is assigned the value zero.


Volume 12 Number 3 March 1981 89










TABLE 11 Effects of husband's occupation on duration of breastfeeding, unadjusted and adjusted through
Multiple Classification Analysis, for wife's age, parity, education, place of residence, and place of work since
marriage, for all women and those who did not use contraception during the last closed birth interval
Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
I. All women
Grand mean, months
of breastfeeding 23.6 19.0 15.7 12.5 11.7 10.0 8.6 8.3
A. Deviations from grand mean (unadjusted)
Husband's occupation
Professional and clerical -3.5 -5.4 -4.8 -3.5 -5.1 -2.1 -3.4 -3.3
Sales and service .7 -2.5 -1.6 .2 -2.1 -1.1 -1.6 -1.3
Skilled and manual .3 -2.6 .8 .4 -1.7 .6 .7 -1.8
Farmers and agricultural .8 3.3 2.0 4.3 3.4 1.5 2.3 4.6
Others .7 -2.4 1.0 1.8 .6 .4 .4 .7
B. Deviations from grand mean (adjusted)
Professional and clerical -1.8 -1.9 -2.8 1.6 -1.9 .5 -1.0 -1.0
Sales and service 0 -1.0 -1.2 .3 .6 .7 -1.0 .5
Skilled and manual .2 -1.3 .4 .4 .6 .5 .5 -1.2
Farmers and agricultural .2 1.4 1.1 2.4 1.2 .8 1.2 2.2
Others .1 -1.4 1.0 1.1 .4 .2 .7 .3
II. Women who did not use contraception
Grand mean, months
of breastfeeding 23.7 19.2 15.6 14.1 12.8 10.6 9.6 9.8
A. Deviations from grand mean (unadjusted)
Husband's occupation
Professional and clerical -3.2 -4.6 -4.3 -3.7 -5.7 -1.4 -4.0 -2.7
Sales and service .6 -1.9 -2.0 .5 -2.8 -1.5 -2.0 -1.4
Skilled and manual .1 -2.7 -1.2 .4 -2.0 .8 -1.1 -2.5
Farmers and agricultural .7 2.6 1.9 3.0 2.5 1.3 1.8 3.4
Others .5 -2.0 .9 .8 .9 .2 .2 .4
B. Deviations from grand mean (adjusted)
Professional and clerical -2.0 -1.6 -2.4 -2.6 -2.5 .2 -1.8 .3
Sales and service 0 .6 -1.5 .5 -1.1 -1.0 -1.5 .5
Skilled and manual .1 -1.6 .8 .2 .7 .6 .9 -1.4
Farmers and agricultural .3 1.2 1.2 2.1 1.0 .7 1.3 1.4
Others .1 -1.4 .9 .7 .2 .3 .3 .7
* = less than 25 cases; 0 = less than .05.


Seven years or more of schooling is assigned a value
of one, and 0-6 years of schooling is assigned the
value zero. Living in urban areas is assigned a value
of one, and living in rural areas is assigned the value
zero. The male child is assigned a value of one, and
the female child is assigned the value zero. "Work"
since marriage is assigned a value of one, and "did
not work" is assigned the value of zero. Women who
did not breastfeed are assigned the value zero for
breastfeeding.)
A simple additive model without any interaction
term is used. The relationship between duration of
breastfeeding and social and demographic factors is
expressed in terms of a constant (intercept), a series
of partial regression coefficients, and an error term.
The salient features of Table 12 are these:


1 The percent variation in the duration of breast-
feeding explained by the seven factors varies
from about 4-5 percent in Guyana and Bangla-
desh to about 27 percent in Peru and 31 percent
in Indonesia.
2 In all countries, the duration of breastfeeding is
shortened if the child dies before reaching one
year of age. This is shown by the negative partial
regression coefficient for infant death.
3 In all countries, women with higher education or
those who live in urban areas breastfed their chil-
dren for shorter periods than others. This is
shown by the negative partial regression coeffi-
cients for education and residence.
4 In all countries, the sex of the child does not im-


90 Studies in Family Planning










TABLE 12 Summary of multiple regression analysis using the duration of breastfeeding (months) as the
dependent variable for all women and for those who did not use contraception during the last closed birth
interval
Independent
variables Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
I. All women
Intercept 24.1 20.4 20.0 13.0 13.7 11.2 7.2 8.7
Age .31* .13* .06 .14* .10* .04 .15* .06
Parity 1.16* .34* .07 .09 .01 .21 .08 .52
Infant death (-14.09*) -17.09* (-13.45*) 9.28* -10.74* (- 5.60*) 8.18* 6.93*
Education of wife 5.99* 6.58* 3.75* 5.47* 5.90* 1.95* 3.67* 3.38*
Residence 2.73* 6.28* 3.14* 3.44* 4.67* 3.10* 2.46* 2.70*
Sex of child .61 .12 .22 .46 .01 .16 .32 .78
Work place of wife .12 1.05 .27 .02 .36 -.12 .01 1.06*
R2 .057 .316 .122 .123 .262 .059 .137 .193
II. Women who did not use contraception
Intercept 23.7 21.0 20.7 13.0 13.3 11.6 7.5 9.2
Age .31* .11* .06 .18* .12* .04 .16* .07
Parity 1.15* .30* .09 .23 .06 .15 .04 .45*
Infant death -13.28* -17.16* -13.20* -10.02* -11.02* 5.72* 8.76* 7.53*
Education of wife 5.38* 6.49* 3.38* 6.34* 5.72* 1.58* 4.16* 2.68*
Residence 2.76* 5.53* 3.52* 2.59* 4.47* 2.86* 2.48* 3.18*
Sex of child .44 .01 .69 .38 .23 .28 .19 .73
Work place of wife .15 .78* .21 .24 .08 .04 .22 .80
R2 .049 .307 .125 .089 .266 .040 .148 .165


( ) Age at death coded differently; figures are not comparable.


ply differential lengths of breastfeeding, even af-
ter adjusting for the effects of the other six
factors. The partial regression coefficients for the
sex of the child indicate that the differences in
breastfeeding between male and female children
are less than one month and are not statistically
significant.
5 Whether or not women worked since marriage
does not have an important and consistent effect
on the duration of breastfeeding.
6 As reported earlier, mother's age and parity do
not show consistent effects on the duration of
breastfeeding. The partial regression coefficients
are either not significant statistically or their
magnitudes are small in comparison to the
effects of such social factors as education and res-
idence.


Social and Demographic
Composition
To what extent can the variations in breastfeeding be-
tween countries be explained by differences in the
social and demographic composition of women? To
answer this question, we have selected wife's educa-
tion, residence, and husband's occupation as the


*Regression coefficient is greater than twice its standard error.


three most important determinants of breastfeeding.
We have shown earlier that differences in the use of
contraception and differences in infant and child
death do not explain the differences in the average
duration of breastfeeding between the eight countries
included in this study. This was indicated by the fact
that the country-specific average duration of breast-
feeding varied to a great extent even among women
who did not use any contraception and among
women whose child was alive at interview (at least
three years after his/her birth). We have also shown
that mother's age, parity, and work place, and the sex
of the child, do not make a significant difference in
the duration of breastfeeding. This leaves mother's
education, her place of residence, and husband's oc-
cupation.
Table 13 compares the observed average duration
of breastfeeding for each country with the estimated
averages. The estimated values for each country are
obtained by using the education-residence or hus-
band's occupation specific averages for that country
and a common distribution of women (obtained by
taking the average for all eight countries). With few
exceptions, the estimated average durations of
breastfeeding are within one month of the observed
averages. These comparisons clearly show that the
observed differences between countries in the aver-


Volume 12 Number 3 March 1981 91









TABLE 13 Average duration of breastfeeding, observed and estimated
Average
duration of
breastfeeding Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
Observed 23.6 19.0 15.7 12.5 11.7 10.0 8.6 8.3
Estimated*
Education and
residence 22.2 17.7 15.8 13.0 12.8 11.3 9.6 10.4
Husband's
occupation 23.4 18.4 15.5 13.8 11.6 10.0 8.6 9.1
*See text.


age duration of breastfeeding cannot be accounted for
by the differences in the social and demographic
composition of women.


Influence of Breastfeeding
on Fertility

The effect of breastfeeding on fertility is suggested by
a number of existing studies in which it is shown
that, in the absence of contraception, the period of
survival of a child is positively associated with the
birth or pregnancy interval in which death occurs.8 It
is assumed that the death of a child truncates the du-
ration of breastfeeding; this in turn leads to an early
resumption of menstruation and ovulation and to an
earlier conception.
A growing body of literature provides more di-
rect evidence for a positive association between dura-
tion of breastfeeding and the length of the birth
interval. A birth interval can be divided into three
main components: (1) postpartum amenorrhea, (2)
menstruating interval, and (3) gestation period. It is
now well established that breastfeeding is the princi-
pal determinant of the duration of postpartum
amenorrhea. In the absence of breastfeeding, menses
return shortly after birth.9 As the duration of breast-
feeding increases, so does the amenorrhea interval-
approximately one additional month of amenorrhea
for each two months' increment in breastfeeding du-
ration. 0 With long lactation, mean amenorrhea inter-
vals from one to two years are observed, in
developing as well as in developed countries."1
A recent analysis of breastfeeding patterns (25
subpopulations from 9 countries in a World Health
Organization Collaborative Study) demonstrated that
after fitting curves with four parameters at any given
time postpartum, variation in breastfeeding propor-
tions explained about 85 percent of the variance be-
tween populations in the proportions of menstruat-
ing women.12 Similarly, other studies have found


high levels of correlation between mean breastfeed-
ing and amenorrhea durations when comparing pop-
ulations13 or subpopulations within countries.14
However, on the individual level, the correlation be-
tween lactation and amenorrhea intervals is lower,
though still highly significant. For example, lactation
explained about 20.7 percent of the variation in the
postpartum amenorrhea periods among Taiwanese
women, which was 92 percent of the total variation
explained by women's age, parity, education, place
of residence, ownership of modern objects, and lacta-
tion.15 The most plausible explanation for the lower
correlation-aside from measurement error-is that
women differ not only with respect to the duration of
breastfeeding, but also with respect to the type and
pattern of breastfeeding.16 It has been demonstrated
that women who breastfeed fully have a lower proba-
bility of resumption of menses than women whose
infants receive supplemental food such as fluids by
bottle or solids.17 The ovulation- and menstruation-
inhibiting effects of breastfeeding, as well as the dif-
ferential impact of breastfeeding types, are believed
to be due to a neurally mediated hormonal reflex sys-
tem initiated by the suckling stimulations of the
breast nipple.'8
There is also some empirical evidence that the
continuation of breastfeeding beyond the resumption
of menstruation suppresses the probability of con-
ception.19 In some societies, breastfeeding is associ-
ated with postpartum abstinence, which, if continued
beyond the resumption of ovulation, will affect the
length of the birth interval independent of the physi-
ological effects of breastfeeding.20
In the present study we will not be able to de-
compose the effect of breastfeeding on the birth inter-
val because the information about the resumption of
menstruation and postpartum abstinence were not
collected in the fertility surveys conducted in the
eight countries included in this analysis. Available
evidence from other studies indicates that the effect
of breastfeeding on the birth interval operates pri-


92 Studies in Family Planning










TABLE 14 Average duration of last closed birth interval by use of contraception in the last closed birth
interval and by duration of breastfeeding for those who did not use contraception
Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
(N=2660) (N=4064) (N=3399) (N=1521) (N=2711) (N=1275) (N=1537) (N=1556)
Birth interval
Mean 38.2 38.4 37.9 35.4 37.6 29.4 35.7 35.3
S.D. 16.9 19.8 20.9 18.3 21.4 18.2 21.9 21.2
Use of contraception
No 38.2 37.6 37.0 32.7 35.6 28.6 31.9 30.7
Inefficient methods 40.3 42.5 43.2 36.5 41.7 33.8 37.9 41.2
Efficient methods 35.8 41.0 39.8 42.3 43.3 32.9 46.0 42.4
Duration of
breastfeeding
(nonusers)
NBF 31.0 33.5 36.7 27.8 32.1 27.7 24.5
0-2 months 36.7 29.7 31.3 33.6 31.8 25.0 29.4 30.5
3-5 39.7 31.0 30.7 28.8 31.6 27.7 27.2 30.3
6 36.3 29.5 33.1 28.5 30.1 29.0 27.7 28.5
7-8 30.1 33.0 32.0 26.4 31.7 26.3 30.2 24.8
9-11 31.2 34.3 29.0 29.7 33.9 29.8 32.7 31.8
12 34.3 35.4 32.5 33.4 35.8 28.3 35.3 31.6
13-17 31.2 33.4 27.5 31.4 33.3 28.6 35.2 29.5
18 33.7 33.4 36.0 33.4 36.7 31.7 36.5 39.3
19-23 38.8 34.6 32.1 32.4 37.8 30.0 33.7 31.5
24 39.9 41.3 42.1 34.8 41.8 32.1 36.8 33.6
25+ 43.8 45.8 54.8 44.7 46.7 34.5 43.6 45.5


marily by delaying the resumption of ovulation after
birth.
In Table 14 we show the average duration of the
last closed birth interval (in months) by use of con-
traception and the duration of breastfeeding. For all
women, the average birth interval is found to vary
from 29 months in Guyana to about 38 months in
Bangladesh, Indonesia, Sri Lanka, and Peru. The
average birth interval was about 35 months in Jordan,
Colombia, and Panama. The use of contraception
generally increases the length of the birth interval;
the magnitude of this increase varies with the country
of residence. Among women who did not use con-
traception, the average length of the birth interval in-
creases with prolonged breastfeeding. For women
who did not use contraception and did not breastfeed
their last-but-one child, the average birth interval is
found to vary from about 24 months in Panama to 37
months in Sri Lanka. These birth interval estimates
are much longer than one would expect in the ab-
sence of breastfeeding and use of contraception.
Other studies have found this interval to be about 20
months.21 The longer birth intervals found here could
possibly reflect differences with respect to average fe-
cundability, temporary separation between spouses,
abortion, and unreported use of contraception; but
they are most likely due to reporting errors.
The effect of breastfeeding on the length of the


birth interval varies between countries. For nonusers,
the differential effect of breastfeeding on the length of
the birth interval is shown by the zero-order correla-
tion coefficients as well as by partial regression coeffi-
cients (see Table 15). On average, one month of
breastfeeding adds about 0.7 months to the birth in-
terval in Sri Lanka; 0.5 months in Indonesia; about
0.45 months in Colombia and Panama; about 0.4
months in Bangladesh and Peru; and 0.3 months in
Jordan and Guyana. This is the net effect of breast-
feeding, after adjusting for the effects of the seven
demographic and social factors included in the mul-
tiple regression analysis. These effects are less than
those found in other studies. For example, in a rural
zone of Senegal, the interval between two births was
found to increase by about nine months for one year
of increase in the age of the child at weaning; that is,
one month of breastfeeding added about 0.75 months
to the length of the birth interval.22 In Taiwan also, it
was found that one month of breastfeeding added
about 0.74 months to the birth interval.23

Breastfeeding,
Use of Contraception,
and Birth Interval

To trace the effects of social and demographic factors
on the birth interval and to assess the relative impor-


Volume 12 Number 3 March 1981 93









TABLE 15 Summary of multiple regression analysis using the last closed birth interval as the dependent
variable for women who did not use contraception during the last closed birth interval
Independent
variables Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
Correlation coefficient .270 .270 .366 .157 .194 .123 .205 .194
Partial regression coefficients
Intercept 35.37 30.22 29.46 33.59 29.35 32.74 27.87 21.23
Breastfeeding .386* .525* .712* .297* .388* .274* .451* .464*
Age .05 .02 .02 .02 .16 .22 .10 .21
Parity .61* .73* .72* .50 .53* .16 .42 .39
Infant death (13.75*) 2.28 (5.70*) 1.58 1.39 (17.93*) .25 1.53
Education of wife 2.66 1.97 1.21 2.53 5.02* 3.28* 3.45 1.09
Residence .11 1.98* .28 .91 .56 .87 1.19 .53
Sex of child 1.16 .01 .11 1.16 .44 1.19 .07 .17
Work place of wife .72 1.53* .14 1.33 1.33 3.92* .57 2.52
R2 .092 .085 .145 .038 .046 .050 .048 .048
( ) Figures coded differently and not comparable.
*Regression coefficient is greater than twice its standard error.


tance of contraception, breastfeeding, and other in-
termediate variables, we used the model depicted in
Figure 2. The arrows to and from other intermediate
variables are shown in broken lines because these
variables are believed to be less important. Using this
model we will test three premises: (1) the length of
the birth interval is primarily determined by the dura-
tion of breastfeeding and the use of contraception; (2)
the effects of other demographic and social factors on
the birth interval are transmitted primarily through
the use of contraception and the duration of breast-
feeding, but could also be transmitted through other
intermediate factors such as fecundability, intra-
uterine mortality, and separation between spouses;
and (3) there is no direct relationship between the use
of contraception and the duration of breastfeeding.
The last premise needs a further explanation. If
breastfeeding is not used deliberately to increase the
interval between two births, but contraception is
used deliberately for this purpose, then the two
forms of behavior should be independent of each
other. In that case, the observed correlation between
the two should be entirely due to their joint associa-
tions with the preceding social and demographic fac-
tors. For example, modernization (as indicated by
mother's education and her place of residence) can
simultaneously result in a decrease in the prevalence
and duration of breastfeeding and in an increase in
the use of contraception. Under these circumstances,
the observed correlation between contraception and
breastfeeding would be spurious. We have shown
earlier that breastfeeding was not used deliberately to


limit the number of children because its duration was
not parity dependent. An empirical test for the third
premise mentioned above will show whether or not
breastfeeding is used deliberately to increase the in-
terval between two births.
In Table 16 we show the regression results for all
women to test the underlying assumptions of the
above model. The correlation and the partial regres-
sion coefficients indicate that there is a positive asso-
ciation between the duration of breastfeeding and
birth interval and between the use of contraception
and birth interval (except for Bangladesh); but the
magnitudes of these effects vary between countries.
Breastfeeding and contraception (and induced
abortion in a few countries) have been shown to be
the most important factors accounting for the differ-
ences between populations in their marital fertility
levels.24 On the individual level it is very difficult to
explain a large proportion of the variance in birth in-
tervals because of the stochastic nature of the re-
productive process. As the regression results indi-
cate, the percent variation in the birth interval
explained by breastfeeding, contraception, and seven
social and demographic factors varies from about 5
percent to 15 percent. A large majority of this ex-
plained variance is due to just two factors-breast-
feeding and use of contraception (compare the values
of R2 in the two regressions with and without social
and demographic factors). This implies that other in-
termediate variables play a small role. This is further
substantiated by the partial regression coefficients for
seven social and demographic factors, indicating that


94 Studies in Family Planning









FIGURE 2 Model to assess relative importance
of contraception, breastfeeding,
and other intermediate variables for fertility


TABLE 16 Summary of multiple regression analysis using the last closed birth interval as the dependent
variable for all women

Independent
variables Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
Correlation coefficients
BI & BF .273 .255 .360 .077 .136 .100 .092 .070
BI & CP .004 .082 .094 .204 .143 .093 .234 .260
BI & Ed .037 .035 .049 .058 .023 .018 .043 .074
BF & CP .008 .025 .012 .250 .173 .170 .157 .210
BF & Ed .116 .250 .184 .236 .321 .170 .201 .337
CP & Ed .294 .229 .183 .317 .258 .166 .273 .266
Partial regression coefficients

Intercept 28.86 29.60 27.32 33.27 33.76 27.36 33.53 33.82
Breastfeeding .395* .465* .674* .168* .327* .212* .250* .174*
R2 .074 .065 .130 .006 .018 .010 .008 .005

Intercept 28.87 28.65 26.47 28.46 30.48 25.96 28.47 27.59
Breastfeeding .395* .469* .672* .298* .399* .253* .359* .323*
Use of contraception .150 4.34* 4.83* 9.17* 7.95* 5.34* 11.68* 12.47*
R2 .074 .073 .138 .059 .047 .023 .072 .084
Intercept 34.93 30.97 29.46 33.74 29.71 30.48 26.77 24.19
Breastfeeding .386* .514* .702* .298* .380* .288* .372* .384*
Use of contraception .01 4.12* 4.58* 8.98* 7.83* 4.85* 11.30* 11.65*
Age .04 .10 .0 .02 .07 .23* .01 .12
Parity .59* .53* .66* .53* .37 .08 .18 .30
Infant death (11.09*) 2.31 (5.56*) 1.04 1.22 (12.30*) .10 .19
Education of wife 1.56 .72 .82 1.28 1.83 2.98* .64 .70
Residence .11 2.84* .16 .30 .48 2.23 .85 .67
Sex of child 1.88 .15 .08 1.28 .57 .08 1.10 .53
Work place of wife .49 1.61* .35 2.01 .82 3.27* .77 3.35*
R2 .090 .085 .147 .069 .050 .046 .074 .093

Partial correlation
between BF & CP,
controlling for other
independent variables .043* .002 .047* .145* .065* .126* .085* .100*
Partial regression coefficient = standard form

Breastfeeding .273 .258 .359 .137 .166 .120 .132 .131
Use of contraception .002 .088 .090 .238 .172 .114 .255 .288

( ) Age at death coded differently; figures are not comparable. *Regression coefficient is greater than twice its standard error.
BI = Birth Interval, BR = Breastfeeding, CP = Use of Contraception, Ed = Women's Education.


Volume 12 Number 3 March 1981 95








in most cases the independent effects of these factors
are either small or are not statistically significant.
The observed negative association between the
use of contraception and the duration of breastfeed-
ing (as indicated by the zero-order correlation coeffi-
cients between the two) is not entirely accounted for
by their joint relationships with the seven social and
demographic factors, such as women's age, parity,
education, and residence. The partial correlation co-
efficients vary in magnitude and direction but are sta-
tistically significant for all countries except Indonesia.
For women in Bangladesh and Sri Lanka, the partial
correlation coefficient is positive; for the remaining
five countries, it is negative. It is possible that this
remaining association is due in part to some other
factors not included in the regression equation and
due in part to reporting errors. The negative partial
correlation coefficients may also indicate that women
in some countries are aware of the fertility inhibiting
effect of breastfeeding and use it for spacing pur-
poses.


Relative Contributions of
Breastfeeding and
Contraception
to Birth Interval

The values of partial regression coefficients of breast-
feeding and use of contraception do not indicate their
relative contributions to the increase in the interval
between two births. These two partial regression co-
efficients are not directly comparable. The coefficient
for breastfeeding indicates the average number of
months added to the birth interval by one month of
breastfeeding. The coefficient for the use of con-
traception indicates the average number of months
added to the birth interval by one user of contracep-
tion. The differences between countries in the partial
regression coefficients for the contraception variable
could be due to differences in the months of con-
traceptive use per user or to differences in the effec-
tiveness of the contraceptive methods used.
The relative contributions of breastfeeding and


TABLE 17 Average number of months added by breastfeeding and use of contraception to the last closed
birth interval
Components of
birth interval Bangladesh Indonesia Sri Lanka Jordan Peru Guyana Colombia Panama
Average number of months
Constant 28.87 28.65 26.47 28.46 30.48 25.96 28.47 27.59
Breastfeeding 9.32 8.91 10.55 3.73 4.67 2.53 3.09 2.68
Contraception 0 0.87 0.87 3.21 2.46 0.96 4.09 4.99
Birth intervals
Estimated 38.19 38.43 37.89 35.40 37.61 29.45 35.65 35.26
Observed 38.2 38.4 37.9 35.4 37.6 29.6 35.7 35.3
Percent distribution
Constant 75.6 74.6 69.9 80.4 81.0 88.1 79.8 78.2
Breastfeeding 24.4 23.2 27.8 10.5 12.4 8.6 8.7 7.6
Contraception 0 2.2 2.3 9.1 6.6 3.3 11.5 14.2
Total 100 100 100 100 100 100 100 100
Estimated median duration of postpartum amenorrhea*
Open birth interval 17.5 14.3 14.0 8.9 8.6 N.A. 5.1 4.7
Closed birth interval 15.2 12.8 10.1 6.6 5.8 4.3 3.5 3.3
Increment in median duration of postpartum amenorrhea
Closed birth interval 13.6 11.2 8.5 5.0 4.2 2.7 1.9 1.7
NOTE: Various components of birth interval are estimated as follows:
BI = a + b (BF) + b2 (CP)
Average Birth Interval = Constant + bi (Average breastfeeding) + b2 (% contraceptive users). The values of the constant 'a' and bl and b2 are
taken from Table 16; the values of average breastfeeding and of % contraceptive users are taken from Table 4.
*Based on Lesthaeghe and Page, cited in note 3. Estimated from Average Duration of Breastfeeding shown in Table 3.


9 Studies in Family Planning








contraceptive use to the interval between two births
are shown in Table 17. The number of months added
by breastfeeding is obtained by multiplying the aver-
age duration of breastfeeding and its partial regres-
sion coefficient. The number of months added by use
of contraception is estimated by multiplying the pro-
portion of women who used contraception and its
partial regression coefficient. These results clearly
show the importance of breastfeeding in extending
the interval between two births at the macro level.
For example, the prevalence of breastfeeding in
Bangladesh, Sri Lanka, and Indonesia added about
9-10 months to the average birth interval. This is
about 25 percent of the length of the birth interval.
The use of contraception in these countries, in com-
parison, added less than one month to the birth inter-
val. In the remaining five countries, breastfeeding
practices added less than five months to the length of
the birth interval, which is 8-12 percent of the aver-
age birth interval in these countries.
A decrease in the prevalence and duration of
breastfeeding would shorten intervals between two
births and, therefore, would increase marital fertility
unless compensated by a simultaneous increase in
the use of contraception. The magnitude of this de-
crease in the length of the birth interval varies from
about 8-10 percent in Colombia, Panama, and Jordan
to about 28 percent in Sri Lanka. We have shown ear-
lier that the duration of breastfeeding is negatively
associated with social indicators. In the absence of
adequate compensation for shortened breastfeeding
by the use of contraception, better educated women,
for example, will have shorter birth intervals than
other women. This is the case in Bangladesh, Indo-
nesia, Sri Lanka, Peru, and Guyana even though
women's education is positively associated with the
use of contraception. In these countries, the zero-
order correlation coefficients between mother's edu-
cation and the length of the birth interval is small but
negative, ranging from -.018 to -.049 (see Table 16).
In the remaining three countries-Jordan, Colombia,
and Panama-the zero-order correlation coefficient
between mother's education and the interval be-
tween two births is positive, ranging from .034 to
.074. In these countries about 35-40 percent of
women used contraception; this contraceptive use
added about 3-5 months to the length of the last
closed birth interval.
Unfortunately, there are good reasons to believe
that some of the figures presented in Table 17 are not
accurate. As already mentioned, the mean duration
of the birth interval, in the absence of breastfeeding
and contraception (estimated by the "constant"), is


substantially higher than the 20 months or so typ-
ically found in other studies. Furthermore, the time
added by breastfeeding is probably substantially
longer in some countries than estimated in Table 17.
This interval should at least equal the increment in
postpartum amenorrhea caused by breastfeeding, be-
cause breastfeeding also may be expected to have
some effect on the menstruating interval. The esti-
mates derived from the regression equation can be
validated by comparing them with independently ob-
tained estimates of increments in postpartum
amenorrhea, using the results of a study by
Lesthaeghe and Page.25 In this study, based on a
large number of data sets, the authors estimated the
expected duration of postpartum amenorrhea for any
duration of breastfeeding up to 30 months. Based on
this relationship, Table 17 gives the median duration
of amenorrhea in the open and closed birth intervals
as well as the increment in postpartum amenorrhea
in the closed birth interval. The average durations of
breastfeeding needed to obtain these estimates were
taken from Table 3. A comparison of the alternative
measure of the breastfeeding effect on the closed
birth interval (last line in Table 17) with the regression
results shows that the latter underestimate the fertil-
ity impact of breastfeeding in Bangladesh, Indonesia,
and Jordan. Whether and to what extent the fertility
impacts of breastfeeding are underestimated in the
remaining five countries is difficult to determine be-
cause general estimates of the effect of breastfeeding
on the menstruating interval are not available. That
the effect of breastfeeding is underestimated in
Bangladesh is further confirmed by studies that have
measured postpartum amenorrhea directly and esti-
mate this interval at about 18 months.26
The discrepancies discussed in the above para-
graph reflect the effect of errors in reporting the ages
of children and the duration of breastfeeding. It is
known that measurement errors in the dependent
and independent variables could bias the estimates of
the constant term and the regression coefficient in the
regression equation. The nature and magnitude of
these biases depend on the mean and variance of the
measurement error and on correlation between the
true value and the measurement error. For example,
assuming that measurement errors and true values
are uncorrelated, any random error in the independ-
ent variable in a regression causes a downward bias
in the regression coefficient.27 This means that any
reporting error in the duration of breastfeeding will
underestimate its impact on the birth interval.
The extent to which observed differences be-
tween countries in the estimated effect of breastfeed-


Volume 12 Number 3 March 1981 97








ing on the birth interval are due to differences in
reporting errors or to differences in other factors-
such as use of contraception-cannot be ascertained
in this study.

Summary

In this paper we have analyzed the patterns of breast-
feeding and its influence on the last closed birth
interval in eight countries: Bangladesh, Indonesia, Sri
Lanka, Jordan, Peru, Guyana, Colombia, and Pan-
ama. The data were taken from the standard recode
tapes made available to the authors by WFS. The re-
sults are briefly summarized below:
1 The large majority of women in all eight coun-
tries breastfed their last two children. The pro-
portion of women who did not breastfeed their
last child ranged from 2 percent in Bangladesh to
18 percent in Panama.
2 The average duration of breastfeeding (including
those who did not breastfeed) varied from nine
months in Panama to about 29 months in
Bangladesh.
3 The key determinants of breastfeeding are:
women's education, place of residence, hus-
band's occupation, and the survival status of the
child. The effects of these factors are consistent
in all eight countries. The results indicate that
women with higher education and those who
live in urban areas breastfeed their children for a
shorter period than those who have lower educa-
tion or live in rural areas. In all countries, the
duration of breastfeeding is shortened if the child
dies before reaching one year of age.
4 The sex of the child does not imply differential
lengths of breastfeeding. Mother's age and parity
did not show consistent effects on the duration
of breastfeeding. Whether or not women worked
since marriage did not show an independent im-
portant effect on the duration of breastfeeding.
5 The differences between countries in the average
duration of breastfeeding are not due to differ-
ences in the composition of women with respect
to the social and demographic factors included in
this study.
6 Breastfeeding is not used for limiting family size,
but we cannot rule out the possibility that it
might have been used to some extent for increas-
ing the interval between births.
7 The average length of the last closed birth inter-
val increased with prolonged breastfeeding in all
eight countries. On average, one month of
breastfeeding adds about 0.4 months to the birth


interval. There is a considerable variation be-
tween countries in the average effect of breast-
feeding. On average, one month of breastfeeding
adds about 0.3 months to the birth interval in
Guyana, Jordan, and Panama; 0.4 months in
Bangladesh, Peru, and Colombia; 0.5 months in
Indonesia; and 0.7 months in Sri Lanka. The
effects of breastfeeding on the birth interval are
underestimated due to reporting errors in the
duration of breastfeeding, especially in Bangla-
desh, Indonesia, and Jordan. Whether and to
what extent the differences between countries in
the fertility impact of breastfeeding are due to the
differences in reporting errors or to differences in
other factors could not be determined.

References and Notes

This paper, originally titled "Socio-biological factors in ex-
posure to child-bearing: Breastfeeding and its fertility
effects," was presented at the World Fertility Survey Con-
ference, London, 7-11 July 1980, and will be published in
the Conference proceedings. The paper is published here
with the permission of the World Fertility Survey.
1 K. Davis and J. Blake, "Social structure and fertility: An
analytic framework," Economic Development and Cultural
Change 4, no. 4 (1956): 211.
2 J. Bongaarts, "The fertility effect of the intermediate fer-
tility variables," paper prepared for IUSSP Seminar on
the Analysis of Maternity Histories, London, April 1980
(to be published in proceedings).
3 Women who had at least one birth were asked how long
they breastfed their last child, unless they were still
breastfeeding at the time of the interview. Taking the
average value of these reported durations of breastfeed-
ing yields a mean that is biased downward, because
women who tend to breastfeed for short periods have a
higher than average chance of being included in the es-
timate.
A simple method exists for obtaining an unbiased
estimate of the mean duration of breastfeeding [J. Bon-
gaarts, "A framework for analyzing the proximate
determinants of fertility," Population and Development Re-
view 4, no. 1 (March 1978): 105-132; R. Lesthaeghe and
H. Page, "The post-partum non-susceptible period: De-
velopment and application of model schedules," Popu-
lation Studies 34, no. 1 (March 1980). H. Page, personal
communication, 1979]. This technique is called the cur-
rent status method because it relies solely on breast-
feeding status at the time of the interview. Let B(t) be
the number of women still breastfeeding at the time of
the interview, among all women who gave birth be-
tween t and t + 1 months before the interview date; and
let N(t) be the number of births that occurred between t
and t + 1 months before the interview date (including
births before that last). The mean duration of breast-
feeding, b, is estimated from:


98 Studies in Family Planning









SB(t)
b = N(t)
0
The upper limit of summation, m, should be set
high enough to cover the longest occurring breastfeed-
ing duration (48 months in this study). The median du-
ration of breastfeeding is given by the month in which
the ratio B(t)/N(t) equals 0.5 (after smoothing as
needed).
4 For example, for Taiwan see Anrudh K. Jain et al., "De-
mographic aspects of lactation and postpartum amenor-
rhea," Demography 7 (1970): 255-271; for Punjab, India,
see R. G. Potter et al., "Applications of field studies to
research on the physiology of human reproduction,"
Journal of Chronic Diseases 18 (1965): 1125-1140; for
women in Bangladesh see L. Chen et al., "A prospec-
tive study of birth interval dynamics in rural
Bangladesh," Population Studies 28, no. 2 (1974):
277-297.
5 See Louis Henry, "Some data on natural fertility,"
Eugenics Quarterly 8 (1961): 81-91; Louis Henry, "La Fe-
condite Naturelle: Observation-Theorie-Resulats,"
Population 16 (1961): 625-636.
6 Jain et al., cited in note 4.
7 See Jain et al., cited in note 4.
8 See, for example, Henry, cited in note 5; Jacques
Henripin, "La Fecondite des Manages Canadiens au
Debut du XVIIIe Siecle," Population 9 (1954): 61-84; J.
Knodel, "Infant mortality and fertility in three Bavarian
villages: An analysis of family histories from the 19th
century," Population Studies 22 (1968): 297-318; Anrudh
K. Jain, "Pregnancy outcome and the time required for
next conception," Population Studies 23 (1969): 421-433.
9 E. J. Salber, M. Feinleib, and B. MacMahon, "The dura-
tion of postpartum amenorrhea," American Journal of
Epidemiology 82, no. 3 (1966): 347-358; J. Pascal, Quelques
Aspects de la Physiologie du Postpartum. These pour le
Doctorat en Medicine, Nancy, 1969; A. Perez et al.,
"Timing and sequence of resuming ovulation and men-
struation after childbirth," Population Studies 25 (1971):
491-503; Chen et al., cited in note 4; P. U. Malkani,
"Menstruation during lactation," Journal of Obstetrics and
Gynecology of India 11 (1960): 11; M. Bonte et al., "Influ-
ence of the socio-economic level on the conception rate
during lactation," International Journal of Fertility 19
(1979): 97; Potter et al., cited in note 4.
10 H. Leridon, Human Fertility: The Basic Components (Chi-
cago: University of Chicago Press, 1977); C. Corsini, "Is
the fertility reducing effect of lactation really substan-
tial?" in Patterns and Determinants of Natural Fertility, ed.
H. Leridon and J. Menken (Liege: Ordina, 1979).
11 Chen et al., cited in note 4; M. Singarimbum and C.
Manning, "Breastfeeding, amenorrhea and abstinence


in a Javanese village: A case study of Mojolama," Studies
in Family Planning 7, no. 6 (June 1976): 175; S. L. Huff-
man, "Nutrition and postpartum amenorrhea in rural
Bangladesh," Population Studies 32 (1978): 251; S. K. Kip-
pley and J. F. Kippley, "The relation between breast-
feeding and amenorrhea: Report of a survey," Journal of
Obstetric, Gynecologic, and Neonatal Nursing 1, no. 4
(1972): 15-21; P. Cantrelle and H. Leridon, "Breastfeed-
ing, mortality and childhood and fertility in a rural zone
of Senegal," Population Studies 25 (1971): 505-533.
12 W. Z. Billewicz, "The timing of postpartum menstrua-
tion and breastfeeding: A simple formula," Biosocial Sci-
ence 11 (1979): 141.
13 Corsini, cited in note 10; Lesthaeghe and Page, cited in
note 3.
14 Salber et al.; Pascal; Perez et al. (all cited in note 9); Jain
et al., cited in note 4.
15 Anrudh K. Jain and T. H. Sun, "Interrelationship be-
tween socio-demographic factors, lactation and post-
partum amenorrhea," Demography India 1 (1972): 3-15.
16 N. Solien de Gonzales, "Lactation and pregnancy: A
hypothesis," American Anthropology 68 (1964): 873; B.
Winikoff, "Nutrition, population and health: Some im-
plications for policy," Science 200 (1978): 895.
17 Perez et al.; Malkani (cited in note 9); Huffman, cited in
note 11; T. McKowen and J. R. Gibson, "A note on
menstruation and conception during lactation," Journal
of Obstetrics and Gynecology of the British Commonwealth 61
(1954): 824.
18 J. E. Tyson and A. Perez, "The maintenance of infecun-
dity in postpartum women," in Nutrition and Human Re-
production, ed. W. H. Mosley (New York: Plenum Press,
1978); P. Delvoye et al., "Serum prolactin in long lasting
lactational amenorrhea," The Lancet 2 (1976): 269.
19 A. Jain, A. Hermalin, and T. H. Sun, "Lactation and
natural fertility," in Patterns and Determinants of Natural
Fertility, ed. H. Leridon and J. Menken (Liege: Ordina,
1979).
20 R. Lesthaeghe and H. Page, Childspacing in Tropical Af-
rica: Traditions and Change (London: Academic Press,
1981, in press).
21 Leridon, cited in note 10.
22 Cantrelle and Leridon, cited in note 11.
23 Jain et al., cited in note 19.
24 Bongaarts (1978), cited in note 3; Bongaarts (1980), cited
in note 2.
25 Lesthaeghe and Page, cited in note 3.
26 A. Chowdhury, "Effect of maternal nutrition on fertility
in rural Bangladesh," in Nutrition and Human Reproduc-
tion, ed. W. H. Mosley (New York: Plenum Press, 1978).
27 J. Johnston, Econometric Methods (New York: McGraw-
Hill, 1972).


Volume 12 Number 3 March 1981 99









Family Size and Sex Preferences

among Women in Rural Bangladesh

Nilufer R. Ahmed


To understand the underlying determinants of high
fertility in a developing country like Bangladesh, the
measurement of preferences for both the number and
the sex of children has become increasingly impor-
tant. Coombs and Sun argue that these preferences
will become more salient as contraceptive use is ex-
tended and results in decreasing the number of un-
wanted births.1 According to them, further changes
in fertility will depend in part on changes in these
preferences. Although Bangladesh has hardly
reached this stage, the assessment of family composi-
tion preferences is useful both because of interest in
present-day correlates and as a point of departure for
examining future changes.
The level of contraceptive use in rural Bangla-
desh is very low. The 1968-69 National Impact Survey
found only 3.7 percent current users of contracep-
tives among married women of reproductive age.
Arthur and McNicoll believe there is no reason to
suppose more than a slight increase in the use of
contraception since the survey,2 although the
Bangladesh Fertility Survey carried out in 1975 esti-
mated about 9.6 percent of currently married women
of reproductive age were current users.3
It has often been argued that high levels of fertil-
ity exist not because there is no demand for con-
traception, but because there is an inadequate supply
and distribution. This contention was tested in 1975
by a saturation-distribution scheme (the Matlab Con-
traceptive Saturation Program) in a rural area of
Bangladesh, where contraceptives were distributed
widely. Results of this program indicate a maximum
of 17-18 percent of current users among married
women of reproductive age. This percentage
dropped to about 12 within two years,4 probably indi-


Nilufer R. Ahmed, M.S., is a graduate student at the Popu-
lation Studies Center of the University of Michigan and
staff demographer at the Bangladesh Institute of Develop-
ment Studies, Dacca.


rating little acceptability of the contraceptives of-
fered. This low level was probably due to a number of
factors, including a lack of motivation coupled with
an underlying preference for large numbers of chil-
dren. Cain believes that contraceptive use in one
rural area is low "because most have no material rea-
son for interrupting the natural reproductive proc-
ess."s
In October 1977 there was a major restructuring
of the Contraceptive Saturation Program, through
which each woman's contraceptive needs were met
with a full range of services. This modified program
had considerably more success, with contraceptive
use of current users increasing to 34 percent 18
months later.6 However, it is doubtful whether such
a program can be replicated on a national scale.
Another argument often given to explain the low
motivation for limiting families in many developing
countries is the strong preference for sons, who are
needed especially for old-age security.7 The 1975
Bangladesh Fertility Survey reveals such a prefer-
ence: of women wanting another child, 62 percent
preferred a boy, compared with 8 percent preferring a
girl; the remaining 30 percent were undecided.
In view of the foregoing it is important to exam-
ine fertility preferences in detail and to investigate the
correlates of such preferences. A step in this direction
is taken in this study of a rural area in Bangladesh.



Data and Sample

The data reported here come from a cross-sectional
two-stage sample of households located in Matlab
than, Comilla, a rural area in central Bangladesh.
Matlab than is an administrative unit with a popula-
tion of about 300,000, covering an area of about 150
square miles. In the first stage of the sample design, a
systematic sample of villages was chosen with proba-
bility proportional to size. In the second stage, four to


100 Studies in Family Planning







five clusters of households were systematically
chosen from each of the sampled villages. This re-
sulted in a total of 1,050 households.
Under the Matlab Contraceptive Saturation Pro-
gram two surveys were carried out: the first (KAP I)
during September-October 1975 and the second
(KAP II) 18 months later. The same 1,050 sampled
households were surveyed both times. Because re-
spondents were limited to currently married women
aged 15-44 years, the sample was not identical in
both surveys. Women who were married since KAP I
were included in KAP II; women who had exceeded
the upper age limit during the interval between KAP I
and KAP II were excluded in KAP II; and so on. In all,
1,246 women aged 15-44 were interviewed: 948 in
both the surveys, 298 in either KAP I or KAP II.
For the purpose of this report, only the 673
women who were interviewed in both surveys, who
were married only once, and who had completed
both the number and sex preference scale questions
(96 percent had done so) are included in the analysis.
This study includes two measures of family size
preferences: a single valued statement of desired
family size and the IN scale measure. Similarly, two
measures of sex composition are included: a single
valued statement of desired sex composition and the
IS scale measure. The scale measures used are based
on a measurement-theoretic approach using conjoint
measurement and unfolding theory.8 These scales,
called I scales, reflect the order of preferences beyond
the first and provide independent measures of pref-
erences for number and sex of children. Respondents
are placed on a psychological continuum ranging
from 1 to 7. The IN scale, which measures preferences
for the number of children, ranges from a small fam-
ily (IN 1) to a large family (IN 7) bias. The IS scale,
which measures preferences for sex of children,
ranges from an extreme girl preference (IS 1) to an
extreme boy preference (IS 7); IS 4 indicates a prefer-
ence for balance of sexes in family composition. It is
believed that these scales reflect an underlying pref-
erence structure that may be more useful in demo-
graphic analyses than previously used global state-
ments (such as desired family size).9 Follow-up
studies have found the scale measures to be better
predictors of subsequent fertility than single valued
statements.10
The information needed to locate an individual's
position on the IN and IS scales is obtained from suc-
cessive choices for family size and composition. In
the long version of the Coombs scale, respondents
are asked to order their preferences of all 16 possible
combinations found in the matrix of 0-3 boys and 0-3
girls, and scale positions based on unfolding theory


are determined. In the short version, respondents are
asked a series of questions to obtain their size and sex
scale positions. Thus, to determine an individual's
family size preference, she is first asked: "If you were
to have the same number of boys and girls, how
many children in all would you most like to have: no
children, two, four, or six?" If the respondent an-
swered "no children," she was in the IN 1 position; if
she answered "six," she was in the IN 7 position. If
she gave any other answer, she was further asked
what her alternate preferences would be if she could
not have that number, and so on, and her scale posi-
tion was determined. An analogous series of ques-
tions was asked to obtain the underlying sex
preferences. The particular questions were not ad hoc
or arbitrarily chosen, but were based on the measure-
ment-theoretic approach taken.'1
In this analysis we examine how number and sex
preferences for children are related to the following
variables: parity, number of living children, and
number of respondent's children who died; age of the
respondents at KAP I and their age at first cohabita-
tion; education of the women and education and oc-
cupation of their husbands; religion; knowledge of,
attitude toward, and use of contraception; perceived
advantages and disadvantages of a large family; de-
sired levels of education for boys and girls; ages at
which sons and daughters are expected to work, ex-
pectations of support from their children, amount of
help expected from their children (not just in old age);
and, finally, expectations of living with their children
sometime in the future. All these variables utilize
KAP I data only. A brief rationale is given below.
Questions have often been raised as to the mean-
ingfulness of inquiries into family size preferences,
especially in developing countries.12 It has been
argued that responses to such questions may often be
meaningless,13 and that respondents often answer
the question by giving the number of children they
already have.14 Hence it is of interest to see how par-
ity or the number of living children is related to re-
sponses to the questions measuring size preferences,
although separating out causal relationships is diffi-
cult.
Nearly all marriages in Bangladesh are arranged
by parents or guardians, and many women are mar-
ried at a very young age. Such women may go to live
with their husband's family but may not have sexual
union with their husbands for a few years. However,
cohabitation often implies sexual union. For this
analysis, age at first cohabitation rather than age at
first marriage is used. A younger age at first cohabita-
tion would mean living with the husband's family at
an earlier age, which might provide a setting for so-


Volume 12 Number 3 March 1981 101








cialization with older traditional ways. Therefore, one
would expect that the younger the age at first co-
habitation, the greater the woman's family size and
the stronger her son preference.
We might expect older women to be more tradi-
tional in their outlook and hence have a stronger
preference for large families and for sons than
younger women, who might be more responsive to
modern attitudes.
Economic necessity together with tradition dic-
tates that in their old age parents live with their chil-
dren, especially sons, or receive at least some help,
usually financial. This combined with the high infant
mortality prevalent in Bangladesh might lead us to
predict that women who had experienced one or
more child deaths may have a stronger preference for
large families and for sons.
Formal education, which usually means some
exposure to more modern views, is expected to be
inversely related to size and sex preferences, while
women with only a religious education (which is very
traditional in its approach) are expected to express
strong preferences for large families and for sons. The
two main religions in the country are Islam and Hin-
duism. More than 85 percent of the population are
Muslims; nearly all of the remainder are Hindus.
Both religions give greater importance to males than
to females, but it is difficult to predict their relative
impact on sex preferences.
Women who perceive greater costs for children
(that is, see only disadvantages in a large family, and
believe that boys and girls should have relatively high
levels of education) are expected to express weaker
preferences for large families. On the other hand,
women who perceive greater benefits from children
(that is, see only advantages in a large family, expect
sons and daughters to work from relatively young
ages, anticipate support from their children in old
age, expect a great deal of help from their children,
and plan to live with their children in the future) are
expected to express stronger preferences for large
families and for sons.



Number Preference
Scale Values

Underlying preferences as measured by the I scale are
different from single valued statements of desired
family size. That is, women giving identical re-
sponses on first preferences may have different un-
derlying values. For example, while two women may
state they would like four children as a first choice for


TABLE 1 Percentage distribution of IN scale values
by stated first preference
Percentage distribution
Small Large
family family
bias bias
Stated family
size preference IN 1-3 IN 4 IN 5 IN 6 IN 7 (N)
1-3 9 37 30 16 8 128
4 2 11 43 25 19 210
5 3 12 17 13 55 110
6+ 2 7 7 7 77 128
Don't know 2 7 11 17 63 97
Total 3 15 25 17 40 673


desired family size, one may be IN 3 and the other IN
6. This indicates that the former woman would rather
have no children than have as many as six, while the
latter would go as high as six rather than have only
two. These two women may be different psychologi-
cally and this difference may affect their fertility be-
havior. Thus the IN (and IS) scales are sensitive to
preferences beyond the first preference, and give a
more accurate picture of what the respondent
wants. 15
The difference between the preference measures,
a single valued statement of desired family size and
the IN scale, is illustrated in the following example.
Table 1 shows that about 19 percent of the respon-
dents say they prefer 1-3 children, a small family in
Bangladesh. Yet only 9 percent of this group had an
underlying preference for a small family (IN 1-3),
while over 50 percent had an underlying preference
for a large family (IN 5-7). Evidence from these IN
scale values indicates that there is a substantial
potential for large families, even among those with a
stated desire for a small family. Thus data on stated
preferences could be misleading in that underlying
preferences are much higher than would be inferred
from single valued statements. It is also of interest to
note that the 97 women who replied "don't know" to
the family size question were able to complete the
paired comparison used to derive the preference
scales. Apparently they found it easier to order an
array of choices than to choose a single most pre-
ferred number.
Results from the study indicate that a very large
majority of women are in the high end of the size
preference scale, so that most of the variations that
exist are between preferences for large and very large
families (Table 1). Eighty-two percent indicated an
underlying preference for large families (IN 5-7), and


102 Studies in Family Planning










TABLE 2 Percentage distribution of IN scale val-
ues by demographic characteristics of wife
Percentage distribution
Small Large
family -- family
bias bias
Characteristic IN 1-3 IN 4 IN 5 IN 6 IN 7 (N)
Age
15-19 2 13 29 19 37 85
20-24 4 12 30 17 37 144
25-29 4 16 23 20 37 142
30-34 3 11 27 20 38 109
35-39 4 20 21 11 44 94
40-44 1 18 17 9 55 99
Age at first
cohabitation
<10 3 17 10 18 51 59
11-12 2 16 24 15 42 104
13-14 3 13 25 18 40 232
15-16 2 14 30 15 38 203
17+ 7 15 29 7 41 41
Parity
0-2 5 13 30 18 34 219
3 1 13 31 15 40 72
4 3 14 19 23 41 73
5 6 15 30 14 35 71
6+ 3 17 19 14 47 238
Number of
living children
0-1 4 12 31 18 35 171
2 3 19 22 22 34 103
3 2 16 29 13 40 97
4 2 13 28 21 35 85
5 6 18 21 15 40 100
6+ 3 14 17 10 56 117
Number of
children who died
0 5 13 28 17 38 354
1+ 2 17 22 16 43 319
Religion
Muslim 3 14 25 17 41 592
Hindu 5 19 27 16 33 81


57 percent have a preference for very large families
(IN 6-7). Only 3 percent have a preference for a small
family, while 15 percent indicated a preference for a
moderate sized family (IN 4).
Table 2 shows that women of all ages have a
preference for very large families. Scale differences
are chiefly between women under and over age 35,
and largely in the IN 7 scale category. About 37 per-
cent of the women under 35 and 55 percent of those
over 40 expressed a preference for very large families
(IN 7), but the younger women showed no greater
preference for moderate sized families (IN 4) than


older woman. Nearly a fourth of those over 35 prefer
a small or moderate sized family, possibly reflecting a
retrospective preference for fewer children than they
have. As will be seen more clearly in the relationships
to parity, some women probably feel that they would
have been better off if they had fewer children; hence
their preference for a moderate or small family.
Perhaps surprisingly, parity is not strongly asso-
ciated with size preferences (Table 2). IN values do
not reflect current family size. Only preference for a
moderate sized family (IN 4) increases systematically
with parity, from a low of 13 percent of women with
two or fewer children to a high of 17 percent of
women with six or more, a difference that disappears
when number of living children is considered. Al-
though IN values are high for most women at each
parity (or number of living children), there is a range
of IN values. For example, a fifth of those with six or
more children would prefer a small or moderate sized
family. It is clear that size preference is not merely a
reflection of how many children the respondent has.
Table 2 further shows substantial differences in
preferences for size between women reporting age at
first cohabitation below 10 years and those reporting
first cohabitation at ages 17 and above. Comparing
these two "first cohabitation" age groups, the data
indicate a decreased likelihood of preference for very
large families among those of later cohabitation age
(from 69 percent to 48 percent on IN 6-7, for the
youngest and oldest age groups respectively), along
with an increased likelihood of preference for some-
what large families (from 10 to 29 percent on IN 5).
However, there is no increase in small family
preferences (IN 1-4). Thus, although an older age at
first cohabitation, with its implication of a later social-
ization with older traditional ways in the husband's
family, is associated with a decreased preference for
very large families, there is no evidence that this
necessarily implies an increased preference for small
families.
Although a high proportion (47 percent) of the
respondents reported at least one child death, per-
sonal experience with infant or child mortality, con-
trary to expectations, has little relationship to
preferences.
Education is negatively related to number prefer-
ences, with nearly 30 percent of those with over six
years of schooling (a very small percent of this sam-
ple) having a moderate to small family size preference
(Table 3). The proportion at the highest end of the
scale drops with increased education. The exception
to this pattern is for women with a religious educa-
tion.
The religious education obtained by Muslim


Volume 12 Number 3 March 1981 103










TABLE 3 Percentage distribution of IN scale val-
ues by selected characteristics of wife
Percentage distribution
Small Large
family 4- family
bias bias
Characteristic IN 1-3 IN 4 IN 5 IN 6 IN 7 (N)
Education of
wife (years)
0 3 20 21 16 40 225
1-5 4 11 35 17 33 164
6+ 11 18 25 14 32 28
Religious 3 12 22 17 46 256
Education of
husband (years)
0 4 19 24 17 36 227
1-5 2 13 24 15 46 178
6+ 5 12 33 15 35 136
Religious 2 13 20 20 45 132
Occupation of
husband
Owns land/
business 3 19 21 15 43 75
Owns/works land 3 12 24 14 47 187
Skilled 5 11 30 15 39 148
Unskilled 1 20 23 15 41 133
Other 5 16 25 25 30 130
Knowledge of
contraception
Yes 4 15 28 17 36 469
No 1 14 20 16 49 204
Attitude toward
contraception
For 5 19 31 16 29 464
Against 0 7 9 17 67 169
Don't know 0 7 20 13 60 40
Use of
contraception
Used 8 23 28 18 23 109
Never used 3 13 25 16 43 564


women includes learning of parts of the Koran and
prayers, and instruction in the beliefs and regulations
of Islam. The teachers are usually local religious lead-
ers, many of whom believe that family planning is
contrary to the teachings of Islam.16 It is possible
therefore that women with such an education would
be more sensitive to the religious issues concerning
family planning and contraception. In fact for such
women number preferences are higher than for any
other group, including those with no schooling at all.
Further, Muslim women are more likely than Hindu
women to have preferences for very large families: 41
percent compared with 33 are IN 7 (Table 2).
There is little systematic relation between hus-
band's education and wife's preference position on


TABLE 4 Relation of IN scale values to use of
contraception
IN scale Percentage ever
values using contraception (N)
1-4 28 123
5 18 169
6 18 111
7 9 270



the IN scale. One would have expected the patterns
of size preference by husband's and wife's education
to be roughly similar. But this is not the case. Hus-
bands with only a primary education (1-5 years) are
more likely to be associated with the wife's prefer-
ence for a very large family (IN 7) than either hus-
bands with no education or wives with a primary
education. Similarly, husband's occupation has little
association with wife's size preference.
Knowledge of, a favorable attitude toward, and
use of contraception have a negative relation to size
preferences. That is, they are associated with a some-
what less extreme preference for large families (Table
3). Especially important in this context is attitude
toward contraception. Roughly 69 percent of the re-
spondents approve of contraception, and 29 percent
of these prefer extremely large families, while 24 per-
cent prefer smaller families (IN 1-4). This finding is in
contrast to 67 and 7 percent respectively among
women who do not approve. One explanation may
lie in the religious view held by some of the respon-
dents that having children is obeying God's will
whereas the practice of contraception is contrary to
God's will. We should note that of those who consid-
ered themselves fertile and did not intend to practice
contraception in the future, almost a third believed
that contraception was contrary to the dictates of
their religion.
Overall, 16 percent of the respondents had used
or were using some form of contraception. Eight per-
cent were current users, a figure that roughly corre-
sponds to that obtained in the Bangladesh Fertility
Survey. The observed negative relation between con-
traceptive use and IN scale values suggests that those
having more moderate preferences may take some
positive step toward achieving their preferences. This
reasoning is reinforced by the observation that at the
lower end of the preference scale (IN 1-4), 28 percent
practiced birth control, whereas only 9 percent of
women preferring large families used some form of
contraception (Table 4).
Perceived costs of children are negatively related
to size preferences. That is, women who believe that
large families (defined in this study to include five or


104 Studies in Family Planning









more children) involve economic and other costs, and
those who desire at least some formal education for
boys and girls have somewhat less extreme prefer-
ences for very large families (Table 5). However,
among women wanting different levels of formal ed-
ucation for children there seems to be little distinction
regarding preferences for large families. But a differ-
ence does appear for preferences for a moderate sized
family (IN 4): women desiring more than 11 years of
education for boys and girls are more likely to prefer a
moderate sized family than women who want lower
levels of education.


In general, perceived benefits of children are
positively related to large size preferences (IN 6-7).
The exceptions (table not shown) seem to be the ages
at which sons and daughters are expected to work;
these show little or no association with IN scale val-
ues. While perceived benefits of a large family are
strongly related to size preferences (only 45 percent
of the women who see no advantages of a large fam-
ily are IN 6-7 compared with over 70 percent who see
some advantages), expectations of support from chil-
dren in old age, of amount of help expected, and of
living with their children are less strongly related.


TABLE 5 Percentage distribution of IN scale values by costs and benefits
Percentage distribution

Small Large
family family
bias bias
Cost variable IN 1-3 IN 4 IN 5 IN 6-7 (N)
Disadvantage of large family
None 1 7 14 77 139
Economic costs 4 16 29 50 470
Other costs 2 20 22 57 51
Desired education for boys
Minimum literacy 4 8 24 64 25
1-10 years 2 5 41 51 41
11-13 years 3 18 25 55 297
College 5 15 26 53 234
Don't know 0 11 16 74 76
Desired education for girls
Minimum literacy 1 9 19 72 81
1-5 years 3 14 32 51 209
6-10 years 5 10 28 57 79
11-13 years 4 20 22 54 197
College 9 21 30 41 47
Don't know 0 8 15 77 60
Benefit variable
Advantage of large family
None 5 18 31 45 387
Economic benefits 1 8 20 72 189
Other benefits 0 15 13 71 53
Expect old-age support
from children
No 0 7 40 53 15
Yes 4 15 25 57 639
Don't know 0 0 36 63 11
Amount of help expected
from children
None 0 17 33 50 6
Little 8 15 27 50 75
Much 3 14 25 58 560
Expect to live with children
No 0 9 36 54 11
Yes 4 15 25 57 647
aAbility to sign name or read letter.


Volume 12 Number 3 March 1981 105









However, the relative strengths of these relations are
hard to guage, because of the very small number of
women who do not expect any kind of help from
their children.



Sex Preference
Scale Values

It is probable that family size preferences are formed
with some consideration of preferences about family
composition (that is, the sex of children). Many past
studies could not disentangle the influences of sex
and size preferences on, say, the future use of con-
traception or the probability of wanting more chil-
dren, since there was no measure of sex bias that was
independent of the number of children desired. It is
important to disentangle these two concepts. For ex-
ample, two women may both say they want three
sons; but if one wants only three children and the
other wants six we cannot assume that their prefer-
ences are equal, as might be done if just the number
of sons wanted was taken into consideration. How-
ever, the IS measure used here is independent of the
number preferences in a measurement sense, so that
a preference for sons can be measured independently
of a preference for large families, and the influences
of each preference on fertility behavior can be sepa-
rated out.
The difference between a single valued state-
ment on desired family composition and the IS scale
is analogous to that between the two preference
measures dealing with family size. The difference is
illustrated in the following example. Table 6 indicates
that about 7 percent of the respondents said they pre-
ferred a family with more girls than boys; another 29
percent stated they preferred an equal number of



TABLE 6 Percentage distribution of IS scale values
by stated first preference
Percentage distribution
Daughter Son
Stated family bias "bias
composition
preference IS 1-3 IS 4 IS 5 IS 6 IS 7 (N)
Boys < girls 5 12 17 34 32 41
Boys = girls 4 6 20 38 32 191
Boys > girls 2 5 18 39 36 338
Don't know 3 1 14 41 41 103
Total 3 5 18 39 36 673


TABLE 7 Percentage distribution of IS scale values
by demographic characteristics of wife: Matlab, 1975
Percentage distribution
Daughter Son
bias bias
Characteristic IS 1-3 IS 4 IS 5 IS 6 IS 7 (N)
Age
15-19 0 2 24 33 40 85
20-24 2 5 13 41 38 144
25-29 3 5 22 33 36 142
30-34 2 4 11 51 32 109
35-39 6 5 19 30 39 94
40-44 6 5 21 41 26 99
Age at first
cohabitation
<10 10 7 13 30 39 59
11-12 2 6 16 37 38 104
13-14 3 1 20 39 36 232
15-16 1 6 18 39 36 203
17+ 2 15 15 44 24 41
Parity
0-2 1 5 18 37 39 219
3 6 6 12 42 35 72
4 5 3 19 33 40 73
5 4 6 25 32 32 71
6+ 3 5 17 42 33 238
Number of
living children
0-1 2 5 18 36 39 171
2 3 4 16 38 39 103
3 6 4 13 40 36 97
4 1 6 24 39 31 85
5 2 8 18 40 32 100
6+ 3 3 20 39 35 117
Number of
children who died
0 1 5 21 38 35 354
1+ 5 4 15 39 37 319
Religion
Muslim 3 4 18 39 35 592
Hindu 2 9 16 36 37 81



boys and girls. Yet, the IS scale shows that over 30
percent of the respondents are either IS 6 or IS 7, in-
dicating a very strong underlying preference for sons.
The single valued statement about preferred sex com-
position can be misleading because it is not inde-
pendent of the number of children wanted and does
not take into account preferences beyond the first.
We would be grossly in error if we took at face value
the statement that over a fourth of the women in
Bangladesh did not prefer sons because they said
they wanted as many sons as daughters. Their un-


106 Studies in Family Planning










TABLE 8 Percentage distribution of IS scale values
by selected characteristics of wife
Percentage distribution
Daughter __ Son
bias bias
Characteristic IS 1-3 IS 4 IS 5 IS 6 IS 7 (N)
Education of
wife (years)
0 4 4 19 35 38 225
1-5 3 5 23 43 26 164
6+ 0 7 25 57 11 28
Religious 3 5 14 36 42 256
Education of
husband
0 3 4 22 35 36 227
1-5 4 6 9 41 40 178
6+ 1 4 24 44 27 136
Religious 4 4 17 36 39 132
Occupation of
husband
Owns land/
business 1 5 15 41 37 75
Owns/works land 5 4 18 35 37 187
Skilled 3 5 18 42 32 148
Unskilled 3 5 14 41 37 133
Other 1 5 24 35 35 130
Knowledge of
contraception
Yes 2 3 19 44 32 469
No 4 8 17 26 45 204
Attitude toward
contraception
For 2 5 21 40 32 464
Against 6 4 14 30 46 169
Don't know 3 5 7 50 35 40
Use of
contraception
Used 3 6 21 56 14 109
Never used 3 4 18 35 40 564


derlying preferences show that 90 percent are on the
son preference end of the scale; few (6 percent) prefer
the balanced family composition that their first stated
preference implies. This is important to recognize,
not only because of implications for fertility if sex
preferences are not satisfied, but because programs of
contraceptive education may be misjudging the situa-
tion if they are based on the single valued statements
alone.
Results from this study (Table 6) show that son
preference is very strong and, indeed, thatmost of
the variation in sex preferences is between strong and
very strong son preferences. The IS scale values indi-


cate that an overwhelming majority, 75 percent, have
a strong son preference (IS 6-7); 36 percent are IS 7,
indicating a son preference so extreme as to exclude a
desire for a daughter, in the categories presented.
Only 5 percent preferred a balanced family composi-
tion (IS 4), while daughter preferences (IS 1-3) are
negligible (3 percent).
The general pattern of strong son preference (IS
6-7) appears among all women, regardless of age,
age at first cohabitation, parity, child mortality expe-
rience, education, religion, education and occupation
of husbands, and knowledge of, attitude toward, and
use of contraception (Tables 7-8). In no subgroup is
there substantial evidence of either girl preference or
preference for a balanced family. However, contrary
to expectations, somewhat less extreme son prefer-
ences are seen among women over 40 (some of whose
preferences for sons may have been modified by hav-
ing had daughters). Women who first cohabited at a
later age, those who had six or more years of educa-
tion, and those who knew and approved of and used
contraception have a somewhat weaker son bias. Es-
pecially important here are women who had used
some form of contraception: 14 percent are IS 7, com-
pared with 40 percent of women who had never prac-
ticed contraception. However, there is little differ-
ence in their preferences for daughters or for a
balanced family (Table 8).
Finally, perceived costs and benefits of children
show little association with sex preferences (tables are
not shown). In each case, as with demographic and
modernization variables, son preferences remain
high.





Conclusion

Underlying preferences for family size among these
rural women are on the high end of the size prefer-
ence scale. A very small proportion of women prefer
a small family or even a moderate sized family. Most
prefer large families, so that variations found in size
preferences are mostly between large families and
very large families. Thus while the relationships be-
tween the IN scale values and various characteristics
of the respondents are generally in the direction ex-
pected, these variations can mostly be seen at the
large family end of the scale. A favorable attitude
toward and use of contraception have a strong nega-
tive relation with number preferences, yet over 40
percent of women in these groups are IN 6-7. Large


Volume 12 Number 3 March 1981 107









family preferences are widely present in all socioeco-
nomic strata and in all age groups. In light of
Coombs's finding in Taiwan that scale measures are
more predictive of future fertility than single valued
statements on desired family size,17 it seems that a
fertility decline in Bangladesh cannot be expected,
even though 50 percent of the respondents have a
desired family size of four or less, unless a change in
underlying preferences is brought about.
Even more pervasive than large family size pref-
erences among these women are strong son prefer-
ences. Almost 93 percent have a son preference; only
5 percent prefer a balanced sex composition. Varia-
tions are therefore even smaller than those found
among size preferences, and are wholly on the high
end of the scale.
Considering that for sex preferences there is very
little differentiation by modernization, it appears as
though the underlying sex preference structure may
be more resistant to change than number preferences
(for which there is a little more variation). This im-
plies that if number preference decreases with mod-
ernization, women would at some point have to
reconcile these two preferences. Most would proba-
bly be unable to achieve their smaller size preferences
and satisfy unchanged sex preferences simulta-
neously. To achieve their desired family composition
they would either have to have more children than
their underlying preference structure indicates, or
have to attain their desired size preference but not
their desired sex preference. The resolution of this
conflict will affect materially the future course of pop-
ulation growth in Bangladesh.
Comparisons with four other developing coun-
tries (Philippines, Taiwan, Korea, and Malaysia) in
which these preference scales have been used'i show
that the women in Matlab have a larger family size
preference as well as a stronger son preference. Fifty-
seven percent of these women are IN 6-7, while 75
percent are IS 6-7. This is in contrast to 48 percent
and 31 percent respectively of Malaysian women, and
24 percent and 54 percent of Korean women.
Finally, it is generally believed that Matlab than
is not very different from other rural areas of the
country, so that the results obtained here may be
generalized to most of rural Bangladesh.



References and Notes

The author is grateful to Raymond L. Langsten, who
was instrumental in collecting the data used in this
paper, for valuable advice; to Lolagene C. Coombs for
helpful suggestions; to James Rogers for program-


ming assistance; and to the International Centre for
Diarrhoeal Disease Research, Bangladesh (Dacca) for
permission to use their data. Support for the analysis
was provided by the Population Studies Center of the
University of.Michigan.

1 L. C. Coombs and T. H. Sun, "Family composition pref-
erences in a developing culture: The case of Taiwan,
1973," Population Studies 32, no. 1 (March 1978): 43-64.
2 W. B. Arthur and G. McNicoll, "An analytical survey of
population and development in Bangladesh," Popula-
tion and Development Review 4, no. 1 (March 1978): 56.
3 Population Control and Family Planning Division,
Bangladesh Fertility Survey: First Report (Dacca, 1978), p.
77.
4 M. Rahman et al., "Contraceptive distribution in
Bangladesh: Some lessons learned," Studies in Family
Planning 11, no. 6 (June 1980): 191-201.
5 M. T. Cain, "The household life cycle and economic
mobility in rural Bangladesh," Population and Develop-
ment Review 4, no. 3 (September 1978): 431.
6 S. Bhatia et al., "The Matlab Family Planning-Health
Services Project," Studies in Family Planning 11, no. 6
(June 1980): 202-212.
7 Cain, cited in note 5, pp. 421-438.
8 A description of this technique can be found in L. C.
Coombs, "Are cross-cultural preference comparisons
possible? A measurement-theoretic approach," Popula-
tion Studies Center Reprint No. 129, University of
Michigan, November 1975. This paper was originally
presented at a conference on the Measurement of Pref-
erences for Number and Sex of Children, sponsored by
the Population Institute of the East-West Center, the
IUSSP, and the Committee for Comparative Behavioral
Studies, June 1975, Honolulu. C. H. Coombs, L. C.
Coombs, and G. M. McClelland, "Preference scales for
number and sex of children," Population Studies 29, no. 2
(uly 1975): 273-298.
9 Coombs and Sun, cited in note 1.
10 L. C. Coombs, "The measurement of family size prefer-
ences and subsequent fertility," Demography 11, no. 4
(November 1974): 587-611; L. C. Coombs, "Prospective
fertility and underlying preferences: A longitudinal
study in Taiwan," Population Studies 33, no. 3 (Novem-
ber 1979): 447-455.
11 Further details of these two versions can be found in L.
C. Coombs, cited in note 8.
12 J. Knodel and V. Prachuabmoh, "Desired family size in
Thailand: Are responses meaningful?" Demography 10,
no. 4 (November 1973): 619-637.
13 P. Hauser, "Family planning and population pro-
grams-a book review," Demography 4, no. 1 (1967):
397-414.
14 R. Bachi and J. Matras, "Family size preferences of Jew-
ish maternity cases in Israel," Milbank Memorial Fund
Quarterly 42, no. 2 (April 1964): 38-56.
15 A note about the tables should be made here. The total


108 Studies in Family Planning








number of respondents for each variable does not al-
ways add up to 673 due to basically two reasons: some
women were assigned a nonexistent code, while some
were not asked the relevant question through oversight.
The size of such discrepancies is small and is believed
not to affect substantially the validity of the results pre-
sented here.


16 M. Franda, "Bangladesh: Perceptions of a population
policy," in Population: Perspective 1973, ed. H. Brown et
al. (San Francisco: Freeman, Cooper and Company,
1973), p. 229.
17 Coombs (1979), cited in note 10.
18 L. C. Coombs, cited in note 8.


Sociocultural Factors and Fertility

in a Rural Nigerian Community


Adefunke Oyemade and Taiwo A. Ogunmuyiwa


A gradual decline in childhood mortality is being
experienced in most developing countries as a result
of improved nutrition and better control of commu-
nicable diseases and parasitic infections. In Nigeria,
educated couples in urban areas are becoming in-
creasingly interested in the possibilities of a fuller life
for their children and are more aware that they have a
better chance of achieving this with a smaller family
size. However, in the rural areas of the country, such
views find less acceptance due to the strong so-
ciocultural values that affect reproductive behavior.'
In this paper, an attempt is made to identify some of
the sociocultural factors affecting fertility in a rural
community in Nigeria.


Material and Methods

This study was undertaken in Badeku, a small village
situated about 27 kilometers north of Ibadan, the cap-
ital of the Oyo State of Nigeria. It has a population of


Adefunke Oyemade, M.D., D.P.H., D.T.H., and Taiwo A.
Ogunmuyiwa, M.B.CH.B., D.P.H., are with the Depart-
ment of Preventive and. Social Medicine, University College
Hospital, Ibadan, Nigeria.


2,394 people living within 275 family compounds.2
The main occupation of its inhabitants is farming. All
family compounds were included in the survey, and
information was collected by means of questionnaires
administered by the authors. Only married women
within the childbearing ages of 15-45 years were in-
terviewed. Questions were asked about their age, ed-
ucation, marriage pattern, fertility, and infant and
childhood mortality. In addition, their opinion was
sought on the ideal number of children per family.

Results

It was not possible to interview a majority of the
women in the family compounds; most of them were
usually out on the farms, while others were engaged
in the local palm oil processing industry, which in-
variably took them outside their homes. In all, 211
women were interviewed, all of whom were found to
be illiterate. Of these, 110 (51.1 percent) were aged
26-35 years, 70 (33.3 percent) were aged 15-25 years,
and only six (2.8 percent) were over age 40.
The women's age at marriage ranged from 14 to
25 years, with a mean age of 17.7. Of the 211 women
interviewed, 189 (89.6 percent) married between the
ages of 15 and 20, 21 (10.0 percent) between ages 21


Volume 12 Number 3 March 1981 109


I









TABLE 1 Marriage patterns and fertility
Average number of
Number of wives Number of women in Total number Average number Average number live births for 10-year
to a man marriage pattern category of live births of live births of years married marriage duration
1 102 298 2.9 9.3 3.1
2 88 288 3.3 11.9 2.9
3 10 34 3.4 13.6 2.5
4+ 11 46 4.1 16.3 2.5
a The standard rate calculated on the simple assumption that number of live births depends linearly on duration of marriage.



TABLE 2 Fertility and childhood mortality
Number of Number of women Average number
childhood deaths with specified number Total number Number of Total number of live births
per woman of childhood deaths of childhood deaths living children of live births per woman
0 130 0 213 213 1.6
1 46 46 111 157 3.4
2 16 32 42 74 4.6
3 6 18 19 37 6.2
4 9 36 21 57 6.3
5+ 4 20 10 30 7.5


and 25, and only 1 (0.4 percent) after age 25. In addi-
tion, 102 (48.3 percent) were monogamously married,
while the remaining 109 (51.7 percent) had married
into polygamous families. Of the latter group, 88 (41.7
percent) were married to husbands with two wives,
10 (4.9 percent) to husbands with three wives, and 11
(5.1 percent) to husbands with four or more wives.
Table 1 shows marriage patterns and fertility: 3.1
live births were recorded among women with mo-
nogamous marriages, 2.9 among women married to
husbands with two wives, and 2.5 among those mar-
ried to husbands with three or more wives.
Table 2 shows the number of live births and
childhood mortality: 1.6 live births was the lowest fig-
ure recorded among women with no childhood
deaths; the number of live births increased from 3.4
among those with one childhood death to 7.5 among
those with five or more childhood deaths.
Among the women sampled, 131 (62.4 percent)
stated that the number of children should be left to
Providence to determine; the average ideal number
stated by the remaining 80 women (37.6 percent) was
8.9.


Discussion

A factor frequently mentioned as contributing to the
changing attitude toward fertility among Nigerian


women is increased education, which has been
shown to be negatively correlated with fertility. All
women in the present study were illiterate, which
may partly account for their high ideal family size. It
is also evident that Nigerian women marry quite
young. The average age of marriage among urban
women in Lagos is 19.8 years, while the average age
recorded in the present study is 17.7 years.
The high proportion of women who thought that
the number of children should be left to Providence
to determine is an indication of the strength of tradi-
tional attitudes toward fertility. The traditional ex-
tended family is bound up with numerous cultural
practices that favor the large family. Under such a
system, the status of women depends upon their
ability to bear children. In rural areas, with their ex-
tensive subsistence agriculture, children are still seen
as indispensable assets in cultivation. Women who
have few children or widely spaced births may be
both pitied and ridiculed.
Finally, the findings indicate that polygamy to
some extent protects the individual woman from ex-
cessive childbearing, although such a marriage sys-
tem places a considerable financial burden on the
husband, who must also provide for the children of
his other wives. Although polygamy is widely prac-
ticed in Nigeria, it is never designed as a method of
birth control. However, controlled fertility could be
initiated in rural areas through the provision of better


110 Studies in Family Planning








educational facilities, improved maternal and child
care services, and, in particular, family planning
services. Considering the traditional values of the
people, a family planning program should be pre-
ceded by an intensive educational campaign covering
the merits of planned parenthood. Such measures
will not only reduce the rural rates of illiteracy, child-
hood mortality, and fertility but will also enhance the
socioeconomic conditions of the community.


References and Notes


1 T. A. Lambo and C. Bakare, "The psychological dimen-
sions of fertility," Rural Africana 14 (1971): 82.
2 A family compound consists of a group of households
numbering from four to six. The most senior male mem-
ber of any household usually becomes the head of the
household and has a great influence on the activities of
household members.


Volume 12 Number 3 March 1981 111









Forum


Health Care for Women
in Latin America and the Caribbean

Mayra Buvinik and Joanne Leslie

Among the most widely used strategies in birth planning has been the designation of a "target
population" of women of reproductive age. Maternal and child health programs are designed as if
this were a uniform category of women, or a functional category for women themselves. In this
paper, Mayra Buvinic, Ph. D. (International Center for Research on Women, Washington, D.C.)
and Joanne Leslie, M.S. (School of Hygiene and Public Health, Johns Hopkins University)
disaggregate women into more meaningful subcategories that look beyond their reproductive role.
What is important about this disaggregation is that it identifies the significant number of women
whose health and family planning needs cannot be met by conventionally designed MCH or family
planning services. Is it sufficient or wise, the authors ask, to design services in effect for only one
group of women-the current childbearers-if these may represent a minority of women in a given
period of time? When subpopulations for health and family planning services are defined along
social, economic, and cultural lines, in addition to reproductive status, striking challenges to the
prevailing design of services for women emerge.


Traditionally, health professionals in Latin America
and the Caribbean, as in most of the developing
world, believe that maternal and child health (MCH)
clinics are appropriate and adequate to meet the
health needs of the majority of women. MCH clinics
continue to provide important services, to women as
well as to children, but as more women delay child-
bearing and/or live well past their childbearing years,
and as women increasingly limit their fertility even
during their childbearing years, it is necessary to look
beyond women's reproductive role in defining their
health needs. Below, we briefly point out some of the
major health, demographic, and economic facts about
women in Latin America and the Caribbean that re-
sult in health needs not met by MCH services. To be
accessible to women and responsive to their health
needs, health delivery systems in the 1980s need to be
designed on the basis of these facts.



Women's Health Status

Mortality
In Latin America and the Caribbean, at all ages above
15, heart disease and malignant neoplasms are among


the five major causes of female mortality. Complica-
tions of childbirth are a major cause of death among
women 15-44 but are not an important cause of death
beyond age 44. Accidents are a major cause of death
of women aged 15-64 but not of women 65 or older.
Beyond age 44, both diabetes and cerebrovascular
disease are major causes of female mortality, and for
women aged 65 and older, influenza and pneumonia
become major causes of death. There are some strik-
ing sex differences in patterns of mortality in Latin
America and the Caribbean, beyond the obvious dif-
ference related to childbirth. For men, accidents are a
major cause of death in all age groups and are the
primary cause of death between the ages of 15 and 44.
Accidents are not the primary cause of female mor-
tality for any age group, and they become a de-
creasingly important cause of death for women in the
older age groups. For men aged 15-44, homicide and
suicide are major causes of death, but these are not
major causes of death for older men or for women in
any age group. Diabetes is a major cause of female
mortality beyond age 44, but it is not a major cause of
mortality for men in any age group.1
An important cause of female mortality in Latin
America and the Caribbean is illegal abortion. There
are few circumstances in which women are legally
entitled to an abortion. However, the available in-


112 Studies in Family Planning








dicators demonstrate high rates of morbidity and
mortality from abortion, as well as an increasing
usage of abortion techniques along with rapid rates of
urbanization in Latin America. Data from both Chile
and Colombia suggest that maternal mortality related
to abortion has increased in the last 20 years. Al-
though the statistics may partially reflect better re-
porting of deaths due to abortion, it appears that in
the 1970s approximately one-third of maternal deaths
in Latin America and the Caribbean were due to in-
duced abortion.2 A study of five cities in Latin Amer-
ica found a greater incidence of abortion among older
than younger women. This finding suggests that
abortion has been used primarily to limit fertility
rather than to space children.

Nutritional Status
Although not a major cause of mortality, the poor nu-
tritional status of many women in Latin America and
the Caribbean contributes significantly to lowered
disease resistance and lowered economic productiv-
ity. Anemia among pregnant and lactating women
has long been recognized as a problem, but nonpreg-
nant women also have much higher anemia preva-
lence than do men. A report on eight studies in Latin
America and the Caribbean showed an average of 36
percent of pregnant women, 15 percent of nonpreg-
nant women, and 4 percent of men to be anemic.3
Low birth weights and high rates of infant mortality
are also indicative of these women's poor nutritional
status. High infant mortality rates are particularly
striking among women under age 20, among women
with no education, among women with five or more
children, and among single mothers.4



Women's Health
Needs

Based on 1977 population figures, we estimate that at
the beginning of 1980 there were 104,180,000 women
over the age of 15 in Latin America and the Carib-
bean. If approximately 30,490,000 of these are preg-
nant or the mothers of children under the age of five
years, there remain 74 million (73,690,000) women
who will not even potentially be reached by the
health services offered under maternal and child
health care programs.5 Who are these 74 million
women, and what are their health needs? On the ba-
sis of age distribution, we estimate that 25,865,000
women are over the age of 45 while 19,847,000 are
teenage women aged 15-19. The rest are women aged
120-45 years who are not pregnant or do not have chil-


dren under the age of five. The following paragraphs
discuss some of the major health needs of different
groups of women who do not fall within the target
population of most MCH programs.

Teenage Women
There are about 20 million women aged 15-19 in Latin
America and the Caribbean; they represent 19 percent
of all adult women. Approximately 13 percent of
births in Latin America and the Caribbean occur to
women 19 years or younger, and about 20 percent of
these occur to teenagers less than 15 years old. The
nutritional requirements of pregnancy and lactation
may be particularly difficult for a teenage woman to
meet, either because these are added to the nutri-
tional requirements for her own growth or because of
the limited purchasing power of many teenage
women. Inadequate nutrition is a major cause of the
low birth weight of infants, which accounts, in part,
for the extremely high mortality rates of infants born
to teenage women in Latin America (over 100 per
thousand live births in many countries). An addi-
tional factor to be borne in mind is that many preg-
nant teenagers and teenage mothers of young
children are unmarried. The social and economic
problems of single mothers are severe, and these are
probably most acute for the teenage single mother.
A large proportion of teenage women in Latin
America and the Caribbean are in institutions of sec-
ondary and higher education, and another large pro-
portion work in domestic service. The health needs of
these two groups of teenage women differ, but both
represent appropriate target groups for family plan-
ning information. Neither group is likely to have rea-
son to visit MCH clinics and both may be fully
occupied during weekday hours; special attention
should be given to alternative ways of making family
planning information and services accessible to such
groups of women.

Older Women
There are about 26 million women aged 45 and older
in Latin America and the Caribbean; they represent
25 percent of all adult women. Only a very small pro-
portion are served by MCH clinics. Most of the health
needs of women in this age group are not related to
their reproductive function, unless one includes un-
der that heading breast cancer and cervical and uter-
ine cancer. The high prevalence of diabetes in this age
group is reflected in the mortality from this disease.
Because the life expectancy of women is greater
than that of men, a substantial fraction of women
over age 45 are widows. These widows, combined


Volume 12 Number 3 March 1981 113







with the women over age 45 who never married,
make up the large population of single older women
in Latin America and the Caribbean (approximately 9
million, or 35 percent of women over age 45). If these
women do not live with an extended family, they
often experience not only the increasing health prob-
lems associated with aging, but also considerable dif-
ficulty obtaining health services due to their limited
mobility and economic circumstances.

Working Women
Any attempt to define strategies for the delivery of
health services to women in Latin America and the
Caribbean must take into account the participation of
women in productive activities. The lack of reliable
statistics on women's economic behavior in develop-
ing countries has resulted in a series of myths that are
now being dispelled. A high proportion of women,
not counted in censuses and conventional labor force
surveys, are employed in low-paid, seasonal occupa-
tions in the rural sector and in the "informal" urban
labor market. Their economic need is reflected in new
data on underemployment and in the increasing pro-
portion of women who are the sole economic pro-
viders for their families. Evidence shows that the
poorest households are headed by women with chil-
dren.6 Many women among the poor in Latin Amer-
ica and the Caribbean have no choice but to work;
planners must look beyond the traditionally acknowl-
edged functions of women as childbearers and child
caretakers, to a realization of women's dual and,
sometimes, conflicting roles as mothers and workers.
As women participate in increasing numbers in the
labor force, particularly in urban industrial jobs, it is
reasonable to expect an increase in occupational dis-
ease and disability among women. In assessing the
health needs of women associated with their place of
employment, it is important to keep in mind that
women may suffer disproportionately from certain
occupation-related diseases or disabilities either be-
cause of physiological differences between men and
women or because women are disproportionately
represented among workers in particular occupa-
tions.

Women-headed Households
Women-headed households-household groups in
which there is no adult male head because he is ab-
sent or not contributing economically-place the
brunt of financial responsibility for survival on
women who are, more often than not, ill equipped to
assume such an obligation. These households are
poor, they are appearing more and more frequently


in economically depressed urban and rural areas, and
female heads are forced into the informal labor sector
in order to support their dependents. In most Latin
American and Caribbean countries for which appro-
priate data are available, we find that a substantial
fraction of single adult women have one or more
child. In Guatemala, 27 percent of single women
have children, in Chile 43 percent have, and in many
of the Caribbean countries more than 50 percent of
single women are mothers. Most of these single
mothers must try to support themselves and their
children. This situation suggests possible needs for
alternative child care for women who are not living in
extended families; protective labor legislation to pro-
vide social security and health care benefits for
women in the informal sector; health care and family
planning services that are easily accessible, in terms
of both location and time; nutrition supplements for
the most economically disadvantaged women; and
subsidized, safe, nutritionally adequate breast milk
substitutes.



Suggestions

A broader understanding of women's health needs
can be gained by examining non-health sector statis-
tics. Data on women's labor force participation, edu-
cational attainments, economic status, and number of
women in each age group can be combined with sex-
specific morbidity and mortality data to provide a
more comprehensive base for health planning to
meet the needs of all women in the 1980s. More spe-
cifically, we suggest the following steps for health
planners:
1 Ensure that sex-specific morbidity, mortality,
and health service data are available. This is es-
sential in order to plan appropriate health serv-
ices for women. Particular efforts should be
made to improve the data available on maternal
deaths due to induced abortion.
2 Develop a functional classification of women and
their health needs. Target groups of women
should be defined in terms of age and reproduc-
tive status, type of economic activity, type of
household (particular note should be made of
women-headed households), and place of resi-
dence (distinguish between urban, rural concen-
trated, and rural dispersed).
3 Promote the integration of health services with
other development activities for women. For ex-
ample, health education could be combined with
literacy or other adult education projects. Also,


114 Studies in Family Planning




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs