Development and Demographic Change:

The Reproductive Ecology of a Rural Ethiopian Oromo Population

Thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy to University College London University of London

Mhalrl Alexandra Gibson

2002 ProQuest Number: U642639

All rights reserved

INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. uest.

ProQuest U642639

Published by ProQuest LLC(2015). Copyright of the Dissertation is held by the Author.

All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code. Microform Edition © ProQuest LLC.

ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 Abstract Across the developing world, labour-saving technologies have been designed to introduce savings in the time and energy that women allocate to work. In rural Arsi, Southern Ethiopia, a water supply pipeline installed in 1996 has reduced long arduous trips spent carrying water, and is associated v^th a permanent improvement in women’s energy budgets. Assuming that nutritional levels remain constant, evolutionary life history theory predicts the energy saved may be diverted into reproductive effort and may increase fertility. Previous clinical studies have revealed that fecundity is responsive to changes in maternal energetic status. This thesis examines whether a reduction in workload translates into a measure of reproductive outcome (live births). The main aims are to detect effects of the installation of village-level taps on birth spacing, and to identify the major correlates of maternal condition, health and child survivorship.

Demographic, health and nutritional status data were collected in 1999-2000, during ten months of field research in Arsi. The survey included a single-round demographic survey with full birth histories and a matemal-child health survey of 1574 households, and an anthropometric survey of 464 reproductive-aged women. These were conducted in villages both with and without access to water points, which were matched for comparability in all other aspects (e.g. altitude, religion, size).

Multivariate event history regression analyses indicate that the introduction of the new water supply scheme is associated with a reduction in first and subsequent birth intervals, following a shorter period of post-partum amenorrhoea. This effect is likely to be mediated by the reduction in women’s energetic workload, as there is no evidence of reduced child mortality or any improvement to women’s health and nutrition. In addition, reproductive scheduling is strongly correlated with maternal condition in this population; evidence for maternal parity-specific depletion in body condition is presented. Furthermore, the Trivers-Willard model of adaptive variation in sex ratio at birth relating to maternal condition is supported.

Overall, the findings indicate that in the absence of adequate health care, food availability and education, the new water supply has not obviously improved maternal well-being or child survivorship. Closer birth-spacing patterns associated with the labour-saving technology may impose severe constraints on maternal (and consequently child) health. These results have potential policy implications for a range of water and other development projects designed to save energy expended by women. Table of Contents

ABSTRACT I TABLE OF CONTENTS I l-V LIST OF TABLES VI LIST OF FIGURES Vli-Vlil ACKNOWLEDGEMENTS IX

1.0 INTRODUCTION 1.1 A im s o f t h e s is 1 1.2 C o n t e x t o f t h e s is 2 1.3 T h e o r e t ic a l a p p r o a c h e s 3 1.3.1 E volutionary l if e h i s t o r y t h e o r y 3 1.3.2 R eproductive e c o l o g y 4 1.4 P a t h w a y s o f f e r t il it y c h a n g e 6 1.4.1 B io -e n e r g e t ic s 6 1.4.2 B io -behavioural f a c t o r s 7 1.5 P a t h w a y s o f m o r t a l it y c h a n g e 8 1.5.1 H e a l t h , f e r t il it y a n d m o r t a l it y interactions 9 1.6 O u t l in e o f t h e s is 10

2.0 METHODS OF DATA COLLECTION AND ANALYSIS 2.1 S t u d y d e s ig n 12 2.1.1 S it e S e l e c t io n 12 2.1.2 S t u d y organisation 14 2.1.3 S a m p l e d e s ig n 15 2.2 L o g is t ic s 16 2.2.1 R e s e a r c h s t a t u s 16 2.2.2 E n u m e r a t io n 17 2.2.3 P il o t & Tr a in in g 17 2.3 H o u s e h o l d demographic s u r v e y 18 2.3.1 Fe r t il it y & m o r t a l it y 18 2.3.2 N u p t ia l it y & r e s id e n c e 19 2.3.4 H o u s e h o l d characteristics 20 2.3.5 Ma t e r n a l s t a t u s 20 2.3.6 W a t e r s u p p l y 21 2.4 B ir t h -s p a c in g d a t a a n d a n a l y s is 21 2.4.1 E v e n t / l if e h i s t o r y c a l e n d a r 22 2.4.2 E v e n t h is t o r y / Ha z a r d s a n a l y s i s 22 2.4.3 M o d e l l in g h a z a r d s 23 2.5 A nthropometric s u r v e y 24 2.5.1 B o d y SIZE 24 2.5.2 B o d y composition 25 2.6 H e a l t h s u r v e y 26 2.6.1 D e f in it io n s o f il l n e s s 26 2.6.2 S e v e r i t y o f il l n e s s 27 2.6.3 H e a l t h c a r e p r a c t ic e s 28 2.7 Q u a l it a t iv e d a t a c o l l e c t io n 28 2.8 D a t a organisation & a n a l y s is 28 3.0 STUDY SITE 3.1 G e o g r a p h y a n d c l im a t e 29 3.1.1 L o c a l e c o l o g y 30 3.2.1 S e a s o n a l v a r ia t io n 30 3.2 E t h n o g r a p h y 31 3.2.1 S u b s is t e n c e p r a c t ic e s 32 3.2.2 V il l a g e /H o u s e h o l d r e s id e n c e 32 3.2.3 Ma r r i a g e p a t t e r n s a n d k in s h ip s t r u c t u r e 33 3.2.4 S o c io -e c o n o m ic c o n d it io n s 35 3.2.5 E d u c a t io n l e v e l s 37 3.3 H e a l t h a n d p o p u l a t io n 37 3.3.1 H e a l t h s e r v i c e s 38 3.3.2 P o p u l a t io n LEVELS 39 3.4 P r o b l e m s o f w a t e r s u p p l y 40 3.4.1 W a t e r c o l l e c t io n 40 3.5 D e v e l o p m e n t infrastructure 42 3.5.1 Th e H it o s a g r a v it y w a t e r s u p p l y p r o j e c t 43

4.0 THE IMPACT OF LABOUR-SAVING TECHNOLOGY ON FERTILITY DIFFERENTIALS 4.1 F e r t il it y l e v e l s 44 4.1.1 P e r io d f e r t il it y r a t e s 44 4.2.1 Fluctuations in f e r t il it y r a t e s o v e r t im e 45 4.2 D o e s l a b o u r -s a v in g t e c h n o l o g y im p a c t o n f e r t il it y ? 46 4.3 M a in a n a l y t ic a l m e t h o d s 47 4.3.1 Da t a s e t 47 4.3.2 D e p e n d e n t v a r ia b l e s 48 4.4 In d e p e n d e n t v a r ia b l e s 49 4.4.1 SOCIO-ECOLOGICAL COVARIATES 49 4.4.2 D e m o g r a p h ic c o v a r ia t e s 50 4.4.3 E n e r g e t ic c o v a r ia t e s 51 4.5 T im in g o f t h e f ir s t b ir t h 52 4.5.1 A g e AT MARRIAGE 53 4.6 A n a l y s e s o f t h e f ir s t b ir t h in t e r v a l 53 4.6.1 Un iv a r ia t e a n a l y s e s 53 4.6.2 M u l t iv a r ia t e a n a l y s e s 55 4.7 A n a l y s e s o f in t e r -b ir t h in t e r v a l s 62 4.7.1 Un iv a r ia t e a n a l y s e s 62 4.7.2 MUL TIVARIA TE ANAL YSES 62 4.8 D is c u s s io n 69 4.8.7 C h a n g e in nutritional s t a t u s ? 69 4.8.2 A l t e r e d b r e a s t -f e e d in g p r a c t i c e s ? 70 4.8.3 R e d u c e d d i s e a s e l o a d s ? 70 4.8.4 R e d u c e d e n e r g e t ic expenditure ? 71 4.8.5 In c r e a s e d c o it a l f r e q u e n c y ? 72 4.9 C o n c l u d in g r e m a r k s 72 5.0 WOMEN'S HEALTH AND BODY CONDITION - EVIDENCE FOR MATERNAL DEPLETION______5.1 T he STATUS OF WOMEN 73 5.2 H ea lth characteristics 74 5.2.1 Health seeking beha viour 75 5.2.2 Reported ILLNESS 76 5.2.3 Miscarriages and still births 76 5.3 L e v e l s o f e n e r g e t ic s t a t u s 77 5.4 D o e s im p r o v e d w a t e r s u p p l y im p r o v e m a t e r n a l c o n d it io n ? 78 5.5 D o e s c l o s e r b ir t h -s p a c in g e f f e c t m a t e r n a l c o n d it io n ? 79 5 . 5.1 Energetic costs of reproduction 80 5.6 A n a l y s e s o f th e m a in determinants o f m a t e r n a l c o n d it io n 81 5.6.1 CORRELATES OF BODY CONDITION 82 5.6.2 C o r r e l a t e s o f h e a l t h 84 5.7 C o n c l u d in g re m a r k s 86

6.0 DETERMINANTS OF MORTALITY 6.1 L e v e l s o f m o r t a l it y 88 6.1.1 P e r io d m o r t a l it y r a t e s 88 6.1.2 Fluctuations in m o r t a l it y o v e r t im e 89 6.2 D o e s im p r o v e d w a t e r s u p p l y im p a c t o n m o r t a l it y ? 91 6.3 A n a l y s e s o f t h e m a in determinants o f m o r t a l it y 92 6.3.1 Y e a r OF BIRTH 93 6.3.2 V il l a g e s it e 94 6.3.3 D e m o g r a p h ic characteristics 95 6.3.4 In d ic a t o r s o f h o u s e h o l d s o c ia l a n d e c o n o m ic s t a t u s 96 6.3.5 S e a s o n o f b ir t h 98 6.3.6 A c c e s s t o w a t e r t a p s t a n d s 99 6.3.7 P r e v io u s b ir t h characteristics 100 6.3.8 Ma t e r n a l b o d y c o n d it io n 101 6.4 C o n c l u d in g r e m a r k s 105

7.0 EVIDENCE FOR ADAPTIVE SEX BIASED PARENTAL INVESTMENT 7.1 S e x b ia s e d in v e s t m e n t t h e o r y 108 7.2 T h e r e l a t iv e c o s t s o f t h e s e x e s 109 7.2.1 Te s t in g a d a p t iv e m o d e l s o f s e x -b ia s e d in v e s t m e n t 110 7.3. C u l t u r a l s e x preferences 111 7.3.1 R e p o r t e d s e x -r a t io b i a s e s 112 7.4 D o e s m a t e r n a l c o n d it io n in f l u e n c e s e x r a t io s ? 114 7.4.1 S e x RATIO AT BIRTH 114 7.4.2 S e x BIASED CHILD MORTALITY 116 7.5 Do SEASONAL PATTERNS INFLUENCE SEX RATIOS? 118 7.5.1 S e a s o n a l f o e t a l l o s s 118 7.5.2 S e x RATIO A T BIRTH 119 7.5.3 S e a s o n a l d i s e a s e p a t t e r n s 121 7.5.4 S e x b ia s e d m o r t a l it y 122 7.6 A r e t h e r e l o n g -t e r m fluctuations in s e x r a t io s ? 124 7.6.1 S e x RATIO AT BIRTH 125 7.6.2 S e x BIASED CHILD MORTALITY 125 7.7 C o n c l u d in g r e m a r k s 129 7.7.1 S u g g e s t io n s f o r Fu t u r e r e s e a r c h 129

8.0 CONCLUSION 131

IV REFERENCES______133-152

APPENDICES______/. Illustrations II. R e g io n a l DATES III. H e a d o f h o u s e h o l d questionnaire IV. E v e r -m a r r ie d w o m a n questionnaire V. H e a l t h questionnaire a n d c h e c k l is t VI. W a t e r u s e Q uestionnaire VII. U s in g e v e n t h i s t o r y c a l e n d a r d a ta VII. T r a in in g m a n u a l s

V List of Tables

2.1 Demographic and socio-ecological characteristics of sample villages 14 3.1 Socio-ecological characteristics of 7 villages in Hitosa and Dodota. 30 3.2 Distribution of indices of household socio-economic status. 36 3.3 Levels of educational attainment for the adult population (>10). 37 3.4 Water carrying workload for women in the dry season months {bonaa). 41 3.5 Description of water-carrying methods in both villages with and without taps. 42 4.1 Data on fertility in Southern Ethiopia. 45 4.2 Variation in mean age at marriage, median age at first birth and length of first birth interval within and between villages since tap installation (March/April 1996). 54 4.3 Univariate median survival time of first birth interval for covariates. 55 4.4 Multivariate hazard regression model for first birth interval (Model 1). 60 4.5 Multivariate hazard regression model for first birth interval with anthropometric covariates. 61 4.6 Description of the major variables for the analysis of length of post­ partum amenorrhoea and open birth intervals. 63 4.7 Multivariate hazard regression model for length of post-partum amenorrhoea and open birth interval (Model 1). 67 4.8 Multivariate hazard regression model for length of post-partum amenorrhoea and open birth interval with anthropometric variables (Model 2). 68 4.9 Median duration of correlates of birth-spacing and mean measures of anthropometric status by village-level access to water tap stands. 68 5.1 Health data from all ever-married women by village. 74 5.2 Indices of village wealth. 75 5.3 Anthropometric data from women in Southern Ethiopia. 77 5.4 Anthropometric characteristics of ever-married women (<50) by village. 78 5.5 Health and nutritional status of women by village water source. 79 5.6 Multiple general linear regression model for correlates of body mass index and sum of skinfolds for reproductive aged women (<50) with over two births. 85 5.7 Multiple logistic regression model for covariates of reported health for all reproductive aged women (<50) with a birth. 86 6.1 Data on early childhood mortality in Southern Ethiopia. 89 6.2 Distribution of births and deaths by each explanatory variable. 102 6.3 Multivariate logistic regression of the probability of dying in three stages of childhood (neonatal, post-neonatal and later childhood). 103 6.4 Multivariate hazard regression model of the probability of dying in early childhood. 104 6.5 Multivariate hazard regression model of the probability of dying in early childhood with preceding birth covariates. 105 6.6 Multivariate hazard regression model of the probability of dying in early childhood with anthropometric covariates. 105 7.1 Multivariate logistic regression of the probability of dying under five by sex of previous sib. 113 7.2 Multivariate logistic regression of the probability of having a male last birth or two measures of maternal condition (a) MUAC, (b) BMI. 115 7.3 Multivariate logistic regression of the probability of the last two births dying in early childhood (<24 months) for (a) maternal MUAC, (b) maternal BMI. 117

VI 7.4 Multivariate logistic regression of the probability of having a male birth by season of birth for all births since 1994. 121 7.5 Multivariate logistic regression of the probability of dying in early childhood (<24 months) by season of birth. 123 7.6 Multivariate logistic regression of the probability of having a male by birth year cohort (1979-2000). 125 7.7 Multivariate logistic regression of the probability of dying in early childhood (<24 months) by birth year cohort (1979-2000). 127

VII List of Figures

3.1 (a) Administrative map of Ethiopia (with Arsi highlighted). 29 (b) The study site; Hitosa and Dodota districts. 29 3.2 Distribution of age at first marriage for all ever-married women. 34 4.1 Total marital fertility rates 1980-1999. 45 4.2 Distribution of female mean age at marriage 1964-2000. 53 4.3 Mean length of completed first birth interval by age at marriage. 57 4.4 Kaplan-Meier plot of the effect of village altitude on probability of first birth. 58 4.5 Kaplan-Meier plot of the monthly probability of first birth for two marital cohorts within villages with access to taps: women married before and after water tap installation in 1996. 59 4.6 Effect of improved access to water points on the timing of return to menses following a birth. 64 5.1 Scatter plot of body mass index by time since last birth. 83 5.2 Scatter plot of sum of two skinfolds by time since last birth. 84 5.3 Mean body mass index by parity and decade of life. 85 6.1 Age-specific mortality for all live births 1979-2000. 90 6.2 Kaplan-Meier plot of the effect of herd size on child survival. 107 6.3 Kaplan-Meier plot of the effect of maternal education on child survival. 107 6.4 Kaplan-Meier plot of the effect of season of birth on child survival. 107 6.5 Kaplan-Meier plot of the effect of maternal BMI on child survival. 107 7.1 Proportion of births dying in childhood (<60 months) by sex of previous sib. 113 7.2 (a) Sex ratio of most recent birth for maternal mid-upper arm circumference. 113 (b) Sex ratio of most recent birth for maternal body mass index. 115 7.3 Proportion of male and female births dying in early childhood (<24 months) for maternal BMI. 117 7.4 Seasonal variation in energy intake (TEI) and expenditure (TEE) for 22 Southern Ethiopian women (redrawn from Ferro-Luzzi et al., 1990). 119 7.5 Sex ratio at birth by season of birth. 120 7.6 Proportions of births dying in early childhood (<24 months) for month of birth by sex. 122 7.7 Cereal producer prices (Ethiopian Birr per metric ton) since 1975. 124 7.8 Trends in sex ratio at birth (1979-2000). 124 7.9 Trends in infant mortality (1969-2000). 128 7.10 Proportion of childhood deaths (<24 months) for sex (1979-2000). 128 7.11 Relative sex ratio of neonatal and postneonatal deaths (1979-2000). 128

VIII Acknowledgements

First, I would like to acknowledge my principal supervisor, Dr Ruth Mace, for introducing me to the topic and providing both inspiration and enthusiasm for my work. I am grateful to Professor Leslie Aiello, Head of Department, for her support throughout my time at UCL. My secondary supervisor Dr Sara Randall deserves thanks for her invaluable advice on collecting demographic data and her useful comments on the final manuscript.

In Ethiopia I am indebted to the staff of WaterAction-Ethiopia & WaterAid, Eshetu Gurmu and Dr Tegegne Gebre Egziabhere of the Institute of Development Studies, Addis Ababa University for kindly helping to facilitate this research. Gratitude is owed to a group of dedicated fieldwork assistants: Demekech Kormie, Hanna Abate, Meseret Beyene, Solomon Worku, Taju Mohammed, Tigist Houndie and Woudinesh Wordofa, for suffering the burden of an increased workload with good humour and friendship. Many other people in Addis and Iteya warrant mention for their kind assistance during data collection. These include Ato Girma Mengistu, Ato Sembeta Lemma and Ato Haile (of WaterAction-Ethiopia), Ato Lemma, Ato Terefa, W/s Askale & Solomon Gebre Selassie. Further, I should like to acknowledge the generous participation of the people of Hitosa and Dodota districts, Arsi and Eastern Showa zones.

Many thanks to the Economic and Social Research Council (UK), the Simon Population Fund and the Parkes Foundation for financially supporting this research.

Within the Anthropology Department at UCL I should like to thank the Reproduction Group for useful comments on early drafts and contributions to lively discussion. Additional thanks are due to Simon Strickland, for advice on conducting the anthropometric measurements, and Rebecca Sear, for statistical help and insightful comments on the final drafts. To the other Darwin Building residents: Nadine Allai, Steve Fraser, Katerina Georgiadis, Clare Holden, Fiona Jordan, Katherine Little, and Andrew Williams: with thanks for 3.5 years of good-humoured co-residence!

Finally, I owe the greatest thanks to my parents, David and Patricia Gibson for continuing to pursue a costly K-selected parental investment strategy.

IX 1

Introduction

1.1 Aims of thesis

In this thesis I shall examine the demographic and health patterns of a rural population in Arsi region, Southern Ethiopia, which has benefited from a labour-saving development technology. The focus is an area of agro-pastoralist subsistence characterised by low food availability, irregular rainfall, poor communication links and limited development infrastructure. Domestic water supply has been highlighted as one of the greatest livelihood problems encountered by the rural community (Silkin 1998). Women typically carry water on their backs and can spend between 3-6 hours a day walking to/from a source of drinking water. However, since 1994 some villages have had the benefit of a gravity-based water development scheme designed specifically to provide village-level access to water. This has greatly reduced the time and effort that women spend on water collection (Silkin 1998).

Access to improved water supply is generally associated with improved matemal-child health and energetic status (White, Bradley et al. 1972; Esrey and Habicht 1986; Yohannes, Streatfield et al. 1992). However, evolutionary life history theory and reproductive physiology predict that a reduction in energy expenditure and an improvement in maternal well-being can lead to higher fertility (Ellison, 1994; Ellison, 2001; Panter-Brick, Lotstein et al. 1993; Gurri, 2001, Pereira et al.). The main aim of this thesis is to investigate whether a water development project, which introduces potentially huge changes in women’s energy budgets, may have unforeseen demographic consequences for a rural community, such as shorter birth-spacing [Chapter 4] and lower child mortality [Chapter 6]. In addition, determinants of fertility, child mortality, maternal body condition and health [Chapter 5] and evidence for sex biasing in parental investment [Chapter 7] are investigated. Chapter 1: Introduction

Traditionally, demographers have sought to describe a number of demographic parameters of a population without necessarily providing an explanatory framework; however, the ultimate theoretical goal of both evolutionary and anthropological demography is a complete explanation of why demographic features are expressed (Pennington and Harpending 1993; Hill and Hurtado 1996; Little and Leslie 1999). Incorporating such an approach, the focus of this thesis will be not only to describe but also to explain any demographic and health patterns arising from the introduction of the new water supply.

1.2 Context of thesis

This thesis contributes to a growing body of bio-anthropological investigation exploring the demographic and health consequences of modernisation among transitional populations (Hill and Hurtado 1996; Little and Leslie 1999). Differential access to the market economy, healthcare facilities, new technologies and relief provisions explain patterns of rural health and population growth across the world (Kunstadter et al. 2001; Kunstadter et al. 1992; Tracer 1994; Tracer 1996).

In most cases, development encompasses a large number of changes in health, economy and social structure, and identifying one outstanding determinant of population change is not always possible. Alterations in dietary intake, activity level, disease and childcare behaviours have multiple effects on different aspects of local demography, some of which are independent, but many of which are almost certain to interact. Problems of endogeneity have impaired previous studies, including distinguishing cause from consequence, and confounding effects of unobserved heterogeneity (Wood 1990). In this thesis the introduction of a single intervention technology, with no community-based initiatives, offers an opportunity to examine a quasi-experimental manipulation of one component of the system - time and energy spent collecting water. By carefully selecting the study site, other confounding effects such as diet, prevalence of disease and other development initiatives are largely controlled, so that the energetic effects associated with a change in women’s workload can be measured independently of these other variables. However, even one such change could influence a number of different domains, from energy budgets to marital practices and health. It is possible to Chapter 1: Introduction disentangle these multiple effects statistically by using high quality data and employing the appropriate statistical models (Chapter 2).

The anthropological literature extensively describes the socio-economic costs and benefits afforded to women following a reduction in workload e.g. by participating in the market economy (Beneria 1982; Carr 1985; Bryceson and Howe 1993; Barrett and Browne 1994). However, the demographic effects of energy and time budget alteration have yet to be investigated, either by academic researchers or the organisations funding these new technologies. Since the direct and indirect effects of population growth present serious socio-economic and health problems for transitional communities in this century, a robust multidisciplinary approach, which considers the impact of development intervention, is vitally important.

1.3 Theoretical approaches

This thesis will combine theoretical models developed in evolutionary and reproductive ecology with demographic data collection techniques to detect population change. The remaining sections of this chapter will introduce relevant theories and methods.

1.3.1 Evolutionary life history theory

Since the landmark studies of Davis and Blake (Davis and Blake 1956) and Henry (Henry 1961), the study of human natural fertility and its determinants has been characterised by a progressive replacement of the Malthusian notion that human fertility is principally regulated by behaviour, with a view that emphasises the regulatory role of biology. To accommodate this theoretical advance, anthropologists have incorporated models developed from evolutionary biology within a larger sociobiological framework in order explain patterns of natural fertility and mortality.

Evolutionary ecology is the study of how environmental and social factors affect the phenotypic pattern of individual populations, starting with the assumption that most variation represents an adaptation to the constraints experienced by the individual. The theory is derived from the neo-Darwinian idea that natural selection has designed Chapter 1: Introduction

organisms to maximise individual reproductive fitness. Evolutionary approaches have been utilised with considerable success to explain the behaviour of other animal species (Krebs and Davies 1978). However, recently evolutionary demographers have sought to test predictions from evolutionary theory in human populations by employing methods and models developed in demography (Bock 1999; Sear 2001; Low, Simon et al. 2002; Strassman and Gillespie 2002).

Life history theory is a branch of evolutionary ecology applied to explain the variation in timing of development, fertility and mortality of living organisms. The central tenet of this theory is the principle of allocation, according to which energy used for one purpose cannot be used for another. As such, a number of trade-offs exist between available energy allocated to growth and maintenance on one hand, and reproduction on the other (Roff 1992; Steams 1992; Low 1993). Life history theory assumes that organisms will resolve these trade-offs in a way that maximises fitness. Human demographic variation is considered to be a function of ecological (including social) context, and adaptive payoffs are cmcial.

Clearly an increased efficiency prompted by a labour-saving technology (such as the water supply scheme) introduces possible savings in time and energy that women allocate to work, which can then be reinvested in child-care, production, domestic work and leisure. Life history theory would predict that if resources are limited and channelled towards either somatic maintenance or reproduction, an increase in labour efficiency should have some effect on reproductive function (providing nutritional levels remain constant).

1.3.2 Reproductive ecology

The field of human reproductive ecology draws heavily from evolutionary life history theory as it is primarily concerned with the way fertility is adjusted to available resources, outlining the importance of environmental constraints on reproductive potential. To date, research in this field has been undertaken largely through clinical studies measuring women’s hormonal levels (Wood 1985; Ellison, Peacock et al. 1986; Bentley 1998). These studies indicate that energetic status impacts on human life history Chapter 1: Introduction traits and reproductive physiology in a number of ways, affecting not only the tempo of growth and maturation (Eveleth and Tanner 1990), but also measures of fecundity, such as hormone levels during cycling (Strassman and Warner 1998; Bentley 1999) and probability of conception (Ellison, Peacock et al 1989; Bailey, Jenike et al. 1992), which mediate the timing of births.

However, endocrine measures alone cannot demonstrate the link between energetic status and fertility as predicted by evolutionary ecology (Tracer, in Utt.). Hormonal studies describe measures of fecundity, which is a theoretical potential, rather than a measure of reproductive achievement. Fertility (a measure of actual live births) and fecundity are obviously related, but the relationship is complex and is mediated by several intervening variables, such as mortality, morbidity and contraception. As a consequence, large differences in fecundity need not translate into major fertility differentials. A study among Bolivian women has revealed that they cycle normally and indeed conceive at salivary progesterone levels significantly below those considered clinically “normal” in the West (Bentley 1998). Additionally, a number of demographic studies have highlighted the importance of early mortality (Hobcraft, McDonald et al. 1983; Alam 1995) and morbidity (Bohler and Bergstrom 1995) in explaining fertility levels.

Consequently, some reproductive ecologists have attempted to collect demographic data in order to assess whether physiological changes have any measurable effects at the population level (Tracer 1991; Little and Leslie 1999), while those researchers with a evolutionary bent have used demographic measures to search for variation in individual fitness (Hill and Hurtado 1996; Strassman and Gillespie 2002). An important reason for using demographic data in the study of reproduction is that they give a precise measure of the phenotypic patterns of a population in the same currency in which natural selection operates. Demographic techniques collect all the relevant parameters for testing important aspects of life history including measures of fertility and mortality. Chapter 1: Introduction

1.4 Pathways of fertility change

Studies on human natural fertility have identified a remarkable heterogeneity in fertility levels both between and within populations. A number of biological and behavioural factors (proximate determinants) have been identified as influencing fertility (Bongaarts and Potter 1983; Bongaarts 1993). Maternal age and energetic factors, which interact with lactation to modulate the period of post-partum infertility, explain the greatest part of the variation (Campbell and Wood 1988; Ellison 1993). However, while age patterns are extremely robust across populations of distinct genetic, ecological and cultural backgrounds, responses to energetic stresses arise as correlates of local ecologies. Nutrition, workload, child care patterns and disease loads all play an important role in explaining levels of natural fertility.

1.4.1 Bio-energetics Numerous clinical studies have revealed that ovarian function varies with the energetic stresses imposed by local ecologies (Bentley, 1985; Bailey, Jenike et al. 1992; Ellison, 2001) and subsequently that patterns of variation in reproductive outcome with energetic factors are general features of human biology. Evolutionary ecology predicts that this variation in response to energetic stress may well represent an evolved feature of women’s reproductive systems (Ellison 1991; Peacock 1991; Ellison 1993). It is a positive response designed by natural selection to enhance lifetime reproductive success by avoiding wasteful allocation of energy to reproductive opportunities with diminished chances of success.

The notion of resource constraint determining levels of fertility has stimulated a considerable amount of research interest into the association between maternal nutritional status and reproduction. A number of studies have attempted to determine whether undemutrition delays reproductive maturation (Foster, Menken et al. 1986), prolongs lactational anovulation and hence intensifies the normal contraceptive effects of breastfeeding (Delgado, Martorell et al. 1982; Huffman 1987), and affects ovarian function, causing elevated risk of anovulatory cycles, oligomenorrhoea or luteal insufficiency (Lager and Ellison 1990; Gumming 1993). Chapter 1: Introduction

One of the earliest studies attempting to demonstrate a relationship between nutritional reserves and fertility was Frisch and MacArthur (Frisch and MacArthur 1974), who stated that there is a critical amount of fat (22%) present in the body which is needed to maintain ovarian function. More recent studies of natural fertility populations have demonstrated that reproductive function is more likely to be constrained by net energy balance (weight change) rather than nutritional energy reserves alone (body fat), (Frisch 1982; Bullen 1985; Lager and Ellison 1990). Ovarian function has been shown to vary with the degree of negative energy balance experienced by populations in diverse ecological settings. For example, among the Lese of Zaire, seasonal suppression of ovarian function corresponded to a significant drop in conception rates, which was most pronounced in the years of greatest weight loss (Ellison, Peacock et al. 1989; Bailey, Jenike et al. 1992). Among the Tamang of Nepal seasonal suppression was recorded among individuals who had lost weight during periods of heavy workload (Panter-Brick, Lotstein et al. 1993).

Other studies indicate that energy expenditure may modulate women’s reproductive function, even at sustained energy balance (Jasienska 2001). Strenuous physical activity suppresses ovarian function (Shangold, Freeman et al. 1979; Beitens, McArthur et al. 1991), thus it is possible that periodic high workloads could account for periods of depressed conception frequencies. In a study among a group of Polish farmworkers, periods of heavy workload were associated with significantly reduced ovarian function, although none of the women were nutritionally stressed or had lost weight (Jasienska and Ellison 1993; Jasienska and Ellison 1998). Moreover, clinical trials indicate that the physiological inhibitory effects associated with high energetic expenditure are reversible following decreasing activity (Gumming, Wheeler et al. 1994).

Such findings suggest that a significant reduction in women’s workload associated with a labour-saving technology may be associated with improved ovarian function and higher conception frequencies, which translate may into reproductive outcome.

1.4.2 Bio-behavioural factors The introduction of a labour-saving technology may also result in behavioural changes, which may directly or indirectly influence the local demography. These include Chapter 1: Introduction

alterations to marital practices, residence patterns and childcare. In a bio-behavioural study of the effects of a grain mill in a rural Mayan population, Kramer and McMillan identified an association with earlier age at marriage for women (Kramer and McMillan 1998; Kramer and McMillan 1999). These authors argue that this was due to changes in the marital decision-making process relating to the role of teen labour. Since teenage girls were no longer required to assist their parents with household chores, they were free to marry at an earlier age. However, an increased rate of biological maturation associated with improved energy budgets could also explain earlier ages at marriage (Gurri, Pereira et al. 2001).

If a reduction in water-carrying workload does free women from time and energy constraints, then this may result in more frequent and more intensive patterns of breast­ feeding; this in turn may reduce ovarian function (even after amenorrhoea has finished) and contribute to lengthening birth intervals (Bongaarts and Potter 1983; Vitzthum, 1989b). However, without clinical data it is difficult to interpret exactly what role is played by lactation in determining fertility in natural fertility populations, since many women resume menses while still nursing and may continue to breastfeed until the next conception. A reduction in workloads may also influence coital behaviour through the interaction of physiological (energetic) factors and improved opportunities for sex.

1.5 Pathways of mortality change Certain researchers have suggested that the ultimate regulator of population levels in traditional populations is not reproduction per se, but mortality (Dyson and Murphy 1985; Wood 1990). As such, it is important to consider potential interactions between fertility and mortality. The demographic literature has extensively described the inter­ relationship between birth interval length and early childhood mortality (Hobcraft, McDonald et al. 1983; Hobcraft 1985; Retherford, Choe et al. 1989; Koenig, Phillips et al. 1990; Alam 1995; Bohler and Bergstrom 1995). However, it is difficult to separate these interactive effects.

Two important interactions between fertility and mortality may be relevant for explaining reproductive outcomes. Firstly, when a nursing child dies, lactation is

8 Chapter 1: Introduction

terminated and the mother resumes ovulating and is exposed to the risk of pregnancy sooner; consequently, high levels of early childhood death can result in increased fertility. Conversely, when a child’s nursing is disrupted by the birth of the next child, the nutritional and immunological benefits are withdrawn and it may be placed at an elevated risk of death (De Sweemer 1984; Wood 1994).

Levels of early childhood mortality may be affected by an alteration in women’s energetic expenditure, specifically by imposing activity-related time limitations on infant feeding patterns (Chen 1974; Huffman, Chowdhury et al. 1980; Panter-Brick 1989; Hawkes 1997). An early introduction of weaning foods or a low frequency and intensity of breastfeeding is associated with reduced immunological and nutritional protection for the child, which ultimately leads to poorer growth and higher infant mortality (Gubhaju, Choe et al. 1987; Lindstrom and Berhanu 2000; Huffman, Zehner et al. 2001). Conversely, if a reduction in water collecting workloads results in improved childcare opportunities for women, then their offspring may benefit from reduced mortality and this in turn may result in wider birth-spacing.

1.5.1 Health, fertility and mortality interactions

Variability in disease prevalence has been used to explain both mortality and fertility patterns throughout the developing world. Access to an improved water supply (quality and quantity) is likely to directly influence morbidity and mortality rates of both women and children by reducing to the prevalence of water-bome/washed diseases (Kloos, Desole et al. 1981; Butz, Habicht et al. 1984; Esrey and Habicht 1986; Esrey, Potash et al. 1991). Furthermore, a general improvement in women’s energy budgets may result in a general improvement in women’s health and well-being (Brabin, Hakimi et al. 2001); conversely, any reduction in disease may improve energy levels (Gilgen, Mascie-Taylor et al. 2001).

A reduced disease burden may affect reproductive function in a non-specific manner analogous to energetic stress (McFalls and McFalls 1984). General ill health has been shown to affect fertility by reducing coital frequency and increasing foetal loss (Mascie- Taylor 1996). Furthermore, common parasitic infections, such as malaria and Chapter 1: Introduction

schistosomiasis, have been associated with pregnancy wastage and tubal occlusion respectively (McFalls and McFalls 1984). Any improvements in women’s health and energetic status may increase fertility through the interaction of the effects of lower workloads and reduced disease loads.

High workloads can also independently influence pregnancy outcome (Bames, Adair et al. 1991; Hatch, Levin et al. 1998). Bio-demographic phenomena which are peculiar to populations experiencing high workloads include an increased risk of foetal loss (Haile 1989), poor intra-uterine growth (Gam 1992) and pre-term births (<37 weeks gestation) (Tafari, Naeye et al. 1980). Babies bom before full term are associated with low birth weights and elevated risk of early infant death (Kline, Stein et al. 1989) and poor health in later life (Barker 1998). A reduction in heavy workloads for women may reduce both intra-uterine mortality and the number of vulnerable pre-term and low birth weight babies. However, these effects cannot be observed in the demographic data when dates of conception are not known.

Overall, an improvement in women’s health and energy budgets associated with the new clean water supply and reduced workload may influence fertility levels through a number of bio-behavioural pathways, including an earlier age at marriage, higher fecundability, increased coital frequencies and more successful full-term pregnancies. Additionally, improved water supplies and child-care opportunities may also reduce the risks of early child mortality. Exploring demographic change associated with new technologies is important since, in the absence of adequate health care facilities, higher fertility and improved child survivorship is likely to increase family size. In a rural area with impoverished resources and limited income generating opportunities, this may place additional demands on women, household resources and local services.

1.6 Outline of thesis Chapter 2 introduces methods of data collection and statistical techniques employed in the analysis of data. Chapter 3 includes a description of the study site and community from which data were collected. Chapters 4-7 are the main data analysis chapters, which address the main aims outlined at the start of this chapter.

10 Chapter 1: Introduction

Chapter 4 examines the main determinants of fertility for the population, specifically addressing the following questions: • Has the introduction of labour-saving technology influenced the timing of reproductive events? • How is this change mediated?

Chapter 5 describes the nutritional and health status of married women, specifically addressing the following questions: • What are the main correlates of maternal health and body condition? • Has the improved access to a water supply influenced maternal condition? • Is higher fertility likely to have an impact on maternal condition?

Chapter 6 explores the main correlates of child mortality, specifically addressing these questions: • What factors predict the risk of early child mortality? • Has the introduction of an improved water supply altered child mortality?

Chapter 7 investigates sex biases in children as predicted by parental investment theory; specific questions addressed are: • Does sex ratio at birth vary according to maternal condition? • Does sex biased child mortality vary according to maternal condition? • Do sex ratio at birth/mortality vary overtime according to resource availability?

Finally, Chapter 8 concludes the thesis by summarising the analyses presented in previous chapters.

11 2

Methods of data collection and analysis

2.1 Study design The focus of the study design was to incorporate a ‘natural experiment’ framework; that is, a study of the same ethnic group living under different conditions. The main difference between these populations relates to the uneven provision of, and accessibility to, a clean water supply. The objectives of the research are to examine both temporal and spatial variation in several demographic and health parameters, as well as to describe the broad demographic characteristics of the villages in the study area.

The main focus is to detect and describe demographic changes: 1. Between villages with and without access to water taps. The variation observed between these villages serving as a proxy for conditions before and after the introduction of modem technology.

2. Within villages across time before and after the arrival of water points. Since the timing of completion for each tap is known, it is possible to seek demographic change at the appropriate time within each village location.

2.1.1 Site selection During the preliminary field trip a number of water development projects were visited in the area south of Addis Ababa. The Hitosa Water Gravity Supply Scheme was chosen as the site most suitable for undertaking the study. Chapter 2: Methods

The fieldwork site was selected using the following criteria: • an area of acute water shortages where women walk long distances on foot to collect water • implementation of a water development scheme which had significantly reduced journey times to collect water • access to both villages benefiting from, and those without access to, water development • an area with no previous history of large-scale development intervention • a settled population • accessibility (160 km from Addis Ababa) • feasibility of working in the area

During the pilot study, seven villages of comparable size and geography were identified within the Hitosa Gravity Water Supply Project's sphere of influence; these included three villages benefiting from water tap installation and four that at the survey date which continued to use the traditional water sources. Villages without taps were selected from those villages in which the water supply pipeline was to be extended. The selection of these study villages permitted sufficient comparability within and between villages, facilitating comparative analyses and the ‘natural experiment’ framework. The demographic composition and characteristics of each village are outlined in Table 2.1.

The nature of this intervention project (Chapter 3: section 3.5.1) involved the laying a pipeline from high to lowland areas through a descending gradient. As such higher and lower altitude villages and those near and far from the original source of water were included within a single installation of the water pipe. The taps in all the villages were turned on simultaneously when the pipeline was completed (in 1996). The timing of the installation of taps does not correlate with altitude of village, proximity to water source or market town. Furthermore, the project was a single intervention, with no community- based activities that might alter other social services such as providing health care or education (Section 3.5.1). However, four model pit latrines have been built in each of the tap villages. Significant in-migration to take advantage of the new water supply is not likely due to the acute shortage of land that prevents new settlement.

13 Chapter 2: Methods

Table 2.1 Demographic & socio-ecological characteristics of the sample villages

Daya Daya Terro Hur Reissa Debula Bekare Debeso Gabrel Moyee -turbe Michiko Saapo Washo All Altitude 2000m 2000m 1960m 1980m 1880m 1800m 1800m

Total pop. 1111 1699 1219 1006 1122 1448 1309 8914

Female 523 838 609 490 533 734 666 4393 Male 588 861 610 516 589 714 643 4521

% under 10 29.25 28.13 30.35 30.81 29.77 29.97 28.65 29.48

EMW 186 279 196 171 211 272 251 1566

EM W age 31.46 32.34 32.52 31.20 30.40 31.93 32.19 31.78 ±SD ± 10.81 ±11.22 ± 10.26 ± 9.90 ± 10.20 ± 10.02 ± 10.69 ± 1 0 .5 5

Anthrop^ 92 86 51 85 88 62 - 464 Household 217 308 182 169 204 247 234 1561

Fem ale led 14.75 15.26 10.99 11.24 10.29 10.53 8.55 11.85 HHs %

Religion % 100 48.7 m 100 100 59.6 m 100 100 70.6 m Christian 51.3 c muslim muslim 40.4 c muslim muslim 29.4 c Econom ic activity^ Maize 1.4 3.5 22.3 8.2 77.3 64.7 87.2 37.6 Wheat/Bar 96.8 94.9 72.3 91.8 17.2 29.3 5.1 58.4 Other 1.8 1.6 5.4 5.4 6 7.7 4

Livestock owned % 79.5 78.3 92.9 94.7 91.1 86.7 89.8 86.7

Date taps installed 1996 1996 2000 2000 None 1996 2000

Latrine use % 10 13.4 --- 2 - 4.4 ^ Sample of ever-married women included in the anthropometric survey ^ Main household economic activity

2.1.2 Study organisation The field study included of 10 months of fieldwork in Arsi region, Southern Ethiopia, employing both demographic and anthropological data collection techniques. The majority of the data were derived from a single-round demographic survey of seven villages in the region and was complimented by measurements of nutritional status collected from a sample of women during both the post-harvest dry seasons. Three main surveys were undertaken:

14 Chapter 2: Methods

Household demographic survey (Appendices III & IV) Demographic data collection was conducted using a single round survey, including a broad household questionnaire and event history calendars for all ever-married women within it. [Appendix I: Illustration 2],

• Household questionnaire This questionnaire recorded the demographic profile, pattern of residence and socio­ economic status and other household data, as well as ascertaining details on water supply and availability. All normally resident household members were recorded, thus providing a cross-section of the household unit and women within it. This survey was divided into two sections; corresponding to questions that were addressed to the head of household (HH) and those answered by reproductive aged (<55 years) ever-married women (BMW).

• Event history calendar for ever-married women Event history calendars were administered to ever-married women, who experienced a birth within the last 6 years, within each household to obtain more detailed timing of reproductive and mortality events over the period that water points were installed in the region. The timing of other significant events, such as water point function, marital status and residence patterns, were also recorded.

Anthropometric and health surveys (Appendix V) Anthropometric measurements and health data were collected from a sample of ever- married women within each village across the fieldwork period in order to assess any changes in health and energetic status following water development. Anthropometric measurements were undertaken during the post-harvest dry season (Dec-Apr) [Appendix I: Illustration 3].

2.1.3 Sample design For the purposes of this study, the primary sampling unit was all reproductive aged (<55 years) ever-married women, within each village. All primary sampling units, within

15 Chapter 2: Methods each village and their household co-members were enumerated in the household survey. With the assistance of village representatives every effort was made to contact and include every household in each village. Household membership was defined as those normally resident (sleeping and eating) within the household compound. A total of 1566 women from 1561 households across the seven villages were enumerated. The total composition of each village population is described in Table 2.1.

A sample of one third of all ever-married women enumerated in the household survey took part in the anthropometric survey during the post-harvest data collection phase. Pregnant and non-reproductive-aged women (>50) were excluded. Ideally a scientifically determined sample population should have been drawn from a complete sampling frame of the villages in the study area. However, due to the lack of census data from the region, the sample women were identified according to logistical considerations, primarily accessibility and availability. Efforts were taken to ensure that a comparable sample of women was selected from both villages with and without water taps (Table 2.1). A total of 464 women were included in the survey.

2.2 Logistics

2.2.1 Research status

During a preliminary field trip (February, 1999) academic links were established with the Demographic Training and Research Centre, Institute of Development Research at Addis Ababa University. Research and residence permits were also obtained from federal, regional and local government authorities.

Water Aid and Water Action staff in Addis Ababa and London facilitated introductions to local community leaders in the villages (Peasant Association Leaders). The associative links, which were established with WaterAid/WaterAction, were vital in ensuring community acceptance and participation. WaterAid/WaterAction has a high profile within the villages, as it remains one of the only aid organisations currently offering humanitarian assistance in the region. Prior to the data collection local village meetings were held with the entire community and key members of the Peasant Association (PA).

16 Chapter 2: Methods

2.2.2 Enumeration

During the demographic data collection phase six field assistants administered the household surveys across the seven villages. They were all identified and tested for levels of comprehension of both English and basic mathematics skills and were educated to secondary school level. Four of the enumerators were recruited from the local area and were familiar with local customs and language dialects (Arsi Oromiffa); while three others were university graduates, recruited in Addis Ababa, who spoke fluent Oromiffa.

The research assistants worked in male-female pairs to complete the surveys in each household. The male enumerator administered the head of household questionnaire, and the female enumerator completed all the ever-married women’s questionnaires. Using same-sex enumerators increased response and accuracy levels, since certain information would have been difficult for an opposite-sexed enumerator to elicit from a respondent, e.g. details concerning menstruation or child death. I undertook all the anthropometric measurements for a sample of ever-married women enumerated with the assistance of one local female field assistant [Appendix I: Illustrations 1, 2 & 3].

2.2.3 Pilot and training

Following the selection and recruitment, the research assistants undertook a two week training session in a village comparable, and adjacent to those selected for study. The training manuals are attached in Appendix VIII. The questionnaire was modified following this pilot trip. The data collected during this pilot are not included in any analyses presented here.

The session included: • an explanation of purpose of study • outline of survey procedures • explanation and discussion of each question and unfamiliar terms in the questionnaire

17 Chapter 2: Methods

• demonstration and practice interviews in the community • discussion and feedback regarding potential problems and possible solutions

2.3 Household demographic survey

The focus of the survey was to describe the major demographic parameters of the study site, and to measure any demographic effects associated with the change in women’s time and energy budgets after the introduction of labour-saving technology. Two techniques for collecting demographic data in the villages were applied; a broad single round household questionnaire in all households and a retrospective birth history and 6 year reproductive events calendar for all ever-married women below the age of 55.

To facilitate data collection within each household, the survey was divided into two sections. • Head of household questionnaire (HH) completed with the head of household, recording information on household composition and the major socio-economic factors that might influence demographic processes (Appendix III). • Ever-married women’s questionnaire (EMW), including a full retrospective birth history, collected from reproductive aged ever-married women within the household. This survey included a more detailed section (event history calendar) concerning the timing of reproductive events (including duration of marriage, births and child deaths) over the 6 years preceding the interview recorded in the event history calendar (Appendix IV).

2.3.1 Fertility & mortality

The topic of childbearing is of interest and significance to both the respondents and the researcher since a woman’s status is related to her fertility as well as to the anticipation of labour assistance and improved old age security. However, the very significance of childbearing can make it difficult to collect complete data, particularly with reference to early child deaths. The reluctance to talk about the dead, especially children, is a major source of bias in fertility surveys (Lucas and Kane 1985). Children who have left home or who have died are often more likely to be omitted. In this study the questionnaire was

18 Chapter 2: Methods

designed by employing standard demographic techniques to ensure data quality (according to the recommendations outlined in (Lucas and Kane 1985)). Data on childbearing and survival were collected from both the head of household and individually from women. Cross-checking the information was possible as the birth histories could be compared with parity sizes reported separately by the head of household. Any discrepancies were addressed at the time of interview.

The birth history was compiled from reproductive aged ever-married women (filtered at <55 years of age), recording information on all births (including miscarriages and still births). The ideal would be to have addressed all women, irrespective of marital status. The argument for this would be that exposure to risk of childbearing is not necessarily confined to women who are married. However, since pre-marital sexual relations were strongly prohibited in the local culture, and marriage was universal for reproductive aged women, only ever-married women were included (Chapter 3: Section 3.2.3).

Each birth history included the sequence of each pregnancy, its outcome, name of the child, age/year of birth, age at death, father’s name and a final question to ensure biological and not adoptive motherhood. Efforts were taken to ensure the accuracy of age reporting; year of birth/death was determined using a list of well-known historical events and, if available, by examining vaccination cards.

2.3.2 Nuptiality & residence

Nuptiality is concerned with the frequency and dissolution of marriages in a population, including the processes of lifetime change in marital status - that is, marriage, divorce and widowhood. It is an important proximate determinant of natural fertility and it may be used as an indirect measure of exposure to the risk of pregnancy (Bongaarts and Potter 1983). The aspect of marriage that is important for this study is the age at which formal sexual union takes place. Age at marriage was considered a reliable determinant of when exposure to childbearing begins, since premarital sex was prohibited. Furthermore, the relationship between marriage system and fertility differentials, such as coital-frequencies, has been well-documented (Dorjahn 1959; Wood 1990). Information on current marital state (including information concerning age at first marriage, number

19 Chapter 2: Methods

of co-wives, current rank) was recorded. A more detailed record of the (monthly) timing of each ever-married woman’s recent marital events, including dates of separation, marital cohabitation and village residence was also recorded in the event history calendar. Additional data concerning village residence was collected separately for each household member.

2.3.4 Household characteristics

Other variables collected in the household questionnaire included: • Age and sex Age and sex were recorded for all those individuals who were named as household members; defined as those who regularly ate and slept within the household compound. • Ethnicity and religion The ethnicity and religion of the head of household was recorded. However, for the purposes of analysis, religion and ethnicity were considered analogous [Chapter 3: Section 3.2]. • Educational level Years of completed education (including primary, secondary, tertiary and adult education) were recorded for each of the household members. Years in education were divided into those at state schools/colleges and those at Koranic school. • Socio-economic status The main economic activity of each household, phrased in terms of time spent in the activity rather than financial returns, was recorded as a crude measure of the subsistence and lifestyle of each household enumerated. Numbers and variety of household owned livestock provided a further indication of household wealth and status. A broad assessment of household’s living conditions was obtained from a series of questions relating to household contents, ownership of material possessions and transportation vehicles, as well as level of sanitation, e.g. use of latrines.

2.3.5 Maternal status

A series of questions in the ever-married women’s questionnaire were directed towards establishing the current maternal status of women. These related to whether, at the time

20 Chapter 2: Methods

of the interview, the woman was known to be pregnant or not, her current breast-feeding status, whether she had returned to menses following a birth and contraceptive use. Ideally lactational status should be measured using breast-feeding frequency and intensity measurements (Vitzthum, 1989a; Vitzthum, 1989b); however, breastfeeding duration relates to length of post-partum amenorrhoea and provides a crude but simple indicator of relative periods of infecundity following child-birth (Nath, Singh et al 1993b).

2.3.6 Water supply

A separate series of questions relating to domestic water supplies were administered all ever-married women (Appendix VI). These questions were directed towards establishing levels of water access and availability for women in the study villages, during both the wet and dry seasons. These questions referred to: • source of drinking water • time taken to collect water/queue for water • quantity of water collected • frequency of water collection • methods of water transportation

2.4 Birth-spacing data and analysis

Women produce a certain number of children by the end of their reproductive lives because of the way in which they time various reproductive events (Wood 1994). Reproduction is a time-dependent process. Models of the birth-spacing process have attempted to study these events and their causes using event history data. An event history is simply defined as a longitudinal record of when events, e.g. births, happen to a sample of individuals (Allison 1984).

Sophisticated statistical methods, such as event history analysis, have been developed for dealing with this kind of duration data (Allison 1984; Yamaguchi 1991) and these have been used to explore the dynamic causal inter-relationships among various aspects

21 Chapter 2: Methods

of birth-spacing. However, these statistical developments have generated a need for detailed life history data which are not available using traditional demographic data collection methods (Hobcraft and Murphy 1986). One approach for utilising the standard cross-sectional survey to obtain additional life history data is by using an event history calendar (Freedman, Thornton et al. 1988).

2.4.1 Event/life history calendar

An Event History Calendar (EHC) is designed to obtain large amounts of information about the timing of events and transitions across many time units. One reason for using such a calendar is that it can improve recall by increasing the respondent’s ability to place different activities/events within the same time frame, thus recording the timing of demographic events as well as data on other more remote influences on reproduction.

Events calendars recorded time-series (monthly) data for births and deaths covering the periods before and after water point installation across all study villages, providing a picture of the demographic profile, including patterns of annual and seasonal change. Calendars were administered to all ever-married women (<55) who experienced a birth within the preceding 6 years, for each household (Appendix IV). This excluded women who would not otherwise be at risk of pregnancy over the study period (e.g. unmarried, infertile and menopausal women). The calendar covered the time period of 6 years prior to the interview month. Thus including the years before and after the initial introduction of water development to the region (1996) as well as a further two-year period, added to avoid any lumping effects during enumeration (1993-2000). Restricting the length of the calendar reduced memory recall bias (Freedman, Thornton et al. 1988). The calendar was divided further into one-month time units, since it was unlikely that respondents would be able to record transitions/events using finer time distinctions. Also, it was unfeasible to fit smaller time units over the 6-year span of the study onto a calendar of manageable size (see Appendix VII on the design and use of calendars).

2.4.2 Event history/ Hazards analysis Event history analysis (EHA) is a method that is employed to analyse time-series data (Tuma and Hannan 1978), as it is concerned with measuring the pattern and correlates of the occurrence of events (Yamaguchi 1991). Unlike standard forms of regression

22 Chapter 2: Methods

analysis, event history analysis has been designed to deal with time-dependent data, e.g. explanatory variables whose value change through time.

Event history analysis also permits the analysis of censored data. One of the greatest problems encountered in using birth interval data obtained from retrospective surveys relates to the incompleteness of women’s maternal histories, the issue of censoring bias. Censoring arises when observations that are begun but not completed during the study, e.g. the current birth interval opened by a woman’s most recent birth. Ignoring this right- censoring bias may lead to the period of exposure being underestimated. Event history analysis is able to incorporate such cases, by including information about their survival up to the time of censoring without making an assumption about the timing of the event’s occurrence in the future (Yamaguchi 1991).

2.4.3 Modelling hazards The principle goal of hazard analysis is to make inferences about the underlying hazard of (risk) from observations on the timing of events. In addition to the risk, it is possible to examine the distribution of times to the occurrence of events (Wood 1994).

In the analysis, discrete-time hazards regression models are estimated to assess the effects of the independent variables, e.g. access to an improved water supply, on demographic events after controlling for other factors. These methods estimate the effect of the predictor variable (e.g. tap installation) on the probability of an event (e.g. birth). The probability is based on the number of events occurring in the risk set, composed of all person months under observation at that time

(Pi = X, p/ (1 + exp (X, P))

X| = vector of covariates for the |th wom an p = vector of regression co efficient

The exponentiated value of the regression co-efficient, exp[p] represents the relative risk of other groups in relation to the baseline group. Exp[p] becomes unity when there is no effect of the covariate, with values greater (or lesser) than unity indicating that the relative risk of an event occurring is greater (or less) for this group than that of the

23 Chapter 2: Methods

reference group.

2.5 Anthropometric survey

In the anthropometric survey, measurements of body size and body composition were collated from a sub-sample of 434 non-pregnant women (<50 years) from six villages during the post-barvest season. These data permit descriptive statements to be made concerning women’s energetic status across all the villages during dry season months, which includes documenting the variation between villages with and without water points, which serve as a proxy for conditions before and after tap installation.

Anthropometry is the comparative study of sizes and proportions of the human body. An evaluation of anthropometric criteria is one of simplest ways to define nutritional and energetic status, which may be calculated with reference to body size and composition (Frisancho 1988; Frisancho 1990). Measurements of weight, height, skinfold thicknesses and upper arm circumference were collated and are used to calculate simple indicators of energetic status and % body fat. Measurements were performed following guidelines set out by Lohman, Roche et al. (1988).

2.5.1 Body size

Measurements of heights and weights of a sample of ever-married women at the time of interview are used to estimate current nutritional status. In adults, body mass index (BMI) is a common simple measure of nutritional status, indicating total body fat (Bailey and Ferro-Luzzi 1995).

BMI (kg/m^) = weight In kg

height' in metres

BMI has been frequently used to diagnose energy deficiency (Ferro-Luzzi, Setter et al. 1992). Chronic energy deficiency is a state in which an adult is actually in energy balance; however it indicates low energy reserves and carries costs in terms of potential risks to health or working capacity. A series of cut-off points have been proposed for evaluating these indices (James, Ferro-Luzzi et al. 1988). While there is no logically

24 Chapter 2: Methods defensible single cut-off point for chronic energy deficiency in adults, a BMI of 18.5 kg/m^ is frequently used as the criteria for normality (James, Ferro-Luzzi et al. 1988).

Limitations • measure of BMI is not a true measure of energy reserves/body fat as it includes an estimate of both fat and lean tissue (Ulijaszek and Strickland 1993). • care must be taken in using tables to estimate CED across populations, as BMI also includes measures of size and shape with current energy reserves, e.g. an Ethiopian woman may survive on low energy intakes, but her weight is also very low (Strickland Pers comm).

2.5.2 Body composition

Body composition is more frequently used to estimate the relative size of different physiological body components, which determine reproductive function (Ulijaszek and Strickland 1993). Percentage body fat has frequently been used as an indication of energy reserves. The most accurate measure of percentage body fat can be obtained by measuring skinfold thicknesses and calculating the percentage body fat from age/sex specific equations based on the sum of skinfolds (e.g. Dumin and Womersley 1974). However, these equations have not been validated across populations (Ulijaszek and Strickland 1993).

In the anthropometric survey, two upper body skinfold measures were recorded (triceps and subscapular) for each women. Variability in thickness of adipose tissues, as indicated by skinfold thickness, is used as a simple indicator of stored energy (% body fat). An additional a measurement of upper arm circumference (MUAC) was undertaken. MUAC can be used to identify levels of peripheral tissue stores of fat and protein (Ferro-Luzzi and James 1996) and provides a practical guide to body muscle mass and nutrition levels (Frisancho 1988).

Limitations • lactating and pregnant women will experience differing levels of body composition to other women

25 Chapter 2: Methods

2.6 Health survey

In order to assess the impact of an improvement in water supply on matemal-child health, a two-week recall health survey (Appendix V) was administered to all ever- married women during the household survey. The information collected was restricted to ever-married women and their young children, identified as the most vulnerable section of the population (Belay 1999), whose health was considered most likely to be affected by improvements in water supply. Women were asked to recall any illnesses that they or their children experienced on the day of survey and during the two weeks preceding the interview. Although such recall questionnaires measure perceived, not actual illness, and are therefore highly dependent on the severity, duration of illness and expectations of health, this method has been recommended for the estimation morbidity levels in the absence of clinical data (Kroeger 1983; Ross and Vaughan 1986).

Limitations • errors due to memory decline bias results • fortnightly interviews produce relative over reporting of recent illness, and under-reporting of past illness, relative to daily interview results; however, Freij and Wall (Freij and Wall 1977) suggest the two effects cancel one another out so the overall reported prevalence of morbidity produces similar results to daily interviews • memory transfer of events from distant to recent past; however this was minimised by specifying recall period clearly

2.6.1 Definitions of illness

A symptoms/conditions checklist developed in a previous health study of the region, presented in the local language was used to aid memory recall and simplify enumeration (Appendix V) (Buschkens and Slikkerveer 1982; Slikkerveer 1990). Responses were coded into the main illness groups during later analysis.

26 Chapter 2: Methods

Limitations • leads to reporting of more minor illnesses involving conditions and symptoms listed • other illnesses not on the list are relatively under-reported

Morbidity rates were calculated as a percentage, reflecting the person’s reporting illness as a percentage of total persons interviewed; these are presented as period or point prevalence rates. Point prevalence rates were determined by asking whether individuals have had any illness or disease on the day of interview. Using this measure captured chronic illnesses and disability that would otherwise be excluded from the analysis (Ross and Vaughan 1986).

2.6.2 Severity of illness

Respondents were asked to estimate the length of each illness/disease episode according to number of days during the recall period. In this survey, events that start prior to the recall period but had extended into it were not included. Accordingly, only disease episodes starting within the recall period were captured within the survey.

Limitations: • difficulties in retrospective allocation of reported symptoms into discrete episodes • problems of incomplete episodes • exclusion of chronic illnesses from the analysis

A further measure of the severity of each episode was obtained from asking respondents about the degree of functional impairment they have suffered due to illness within the recall period. The severity of illness was judged on whether the respondent was ill to a degree that he/she was required to lie down during the day.

27 Chapter 2: Methods

2.6.3 Health care practices

Information regarding local health care practices and the availability of modem medicines was obtained from questions regarding the treatment of each bout of illness. These data were used to provide some indication of changes in health-seeking behaviours. A further question on the uptake and timing of childhood immunisation was included in the survey.

2.7 Qualitative data coliection

Throughout the ten-month fieldwork period, participatory data collection techniques were employed, in order to investigate the opinions of local community members and regional specialists relating to: • the costs and benefits for the community of the new water point installation • water point function and reliability • birth-spacing beliefs and practices • family planning and health care access/use • childcare practices and sex preferences • major health concerns for women and children

These included focus-lead discussion groups with same-sex village-members and semi­ structured and/or unstructured interviews with key members of the local community, e.g. regional health care professionals, government and NGO workers, community leaders and tap attendants.

2.8 Data organisation and anaiysis

The collected data were coded, compiled and organised using Access and Excel computer software. In the following chapters both simple descriptive statistics as well as other hypothesis-testing statistics are employed to analyse the data using Statistical Package for the Social Sciences (SPSS version 9.0) and Statistical Analysis System (SAS version 6.1) statistical software.

28 3

Study Site

‘...to the Abyssinian women falls the hardest of the communal tasks throughout a life which prematurely ages and destroys her. She Is the hewer of wood and the drawer of water, and as the villages are always distant from either she tolls for miles with Incredibly heavy weights on her back’ (Forbes 1925)

3.1 Geography and climate The focus of this thesis is seven villages in Arsi Administrative Zone, Oromiya region of Southern Ethiopia. The villages are located in the Hitosa and Dodota woredas (districts), within 10km of a small market town, Iteya, and 20 km north-east of Asella, the regional capital (Fig. 3.1 & Table 3.1). Temperatures range from 9.8°C to 21°C (Cohen and Isaksson 1987). Rainfall levels are low, around 700-900mm/annum (Gamachu 1977). With the exception of the rainy season, (June to September), the climate is very dry with occasional rain.

Fig. 3.1 a) Administrative map of Ethiopia (Arsi highlighted) b) Study site: Hitosa & Dodota

PD^. (SOUTH V g A d d is Ababa j fiGRAY \ \ SHEWA y I :A.

Djibouti 1 y y I N ^ e b r e Zeyit / r ,N G Ç JA M

‘-'.WKUGA SOMALI/ t " ' j T 1 ! A flrÿO W^fJî \ y A çK'ok’a -€ëSîi/ HARERGE

{ ^ b A L L L • 1 n .f , ' . U ..Ne««S" _ ' "'(slDAMO n Asela Lafiè Zway i; ) SOMALIA _ . ) / \ ( Chapter 3: Study site

3.1.1 Local ecology All the villages in the study area lie within thebaddadaree (semi-highland climate zone), which refers to all lands between 1400-2000 m.a.s. To facilitate the analysis the villages have been broadly categorised as being located in either lowland (<1900 m.a.s.), or highland (>1900 m.a.s.) areas (Table 3.1). In the lowland villages the climatic zone is characterised by unreliable rainfall (<800 mm/year), high temperatures (min. 13°C to max. 27°C), extreme scarcity of water and soil erosion. This area is of the dry savannah vegetation type with scattered trees (acacia), large shrubs and perennial grass, which is highly suitable for grazing cattle [Appendix I: Illustration 4].

In the highland area, there is heavier rainfall (800-900 mm/year), cooler temperatures (min. 10°C to max. 23 °C), which provides suitable environmental conditions to support a wider variety of crop types, e.g. barley, teff and leguminous vegetables.

Table 3.1 Socio-ecological characteristics of the villages in Hitosa and Dodota (total population= 8914)

Village N Altitude Main Journey time Journey time Access (m.a.s) religion to collect water to collect water to before 1996* after 1996* Taps** Daya 1111 High Christian 6 hours <30 minutes Yes Debeso 2000m Daya 1699 High Christian/ 6 hours <30 minutes Yes Gabrel 2000m Muslim Terro 1219 High Muslim 4-6 hours 4-6 hours No Moyee 1960m Hurturbe 1006 High Muslim 4-6 hours 4-6 hours No 1980m Reissa 1122 Low Christian/ 3 hours 3 hours No M ichlko 1880m Muslim Debula 1448 Low Muslim 6 hours <30 minutes Yes Saapo 1800m Bekare 1309 Low Muslim 3 hours 3 hours No W asho 1800m Notes: m.a.s.= metres above sea level n= resident population in week of survey * during height of dry season (Dec-Apr); local ground water is collected in wet months ** at date of survey

3.2.1 Seasonal variation The local calendar is based on four agricultural seasons. Biraa (Sept-Nov), the main harvest month, is the start of the Oromo year and is described as the ‘morning of each year’, which is associated with childhood and happiness. During the first months of

30 Chapter 3: Study site biraa early crops such as wheat and barley are harvested and milk products are readily available. The next season is bonaa (Dec-Feb) when there is a progressive rise in temperature. Rituals such as circumcision and marriage take place at this time of the year (Terefe 2000). Rain begins to fall again in arfassa (Mar-May), which is the season farmers begin to prepare their land and sow seeds for the next year’s harvest. Around the end of this season the food supplies begin to diminish. The highest rainfall occurs during ganna (Jun-Aug) and temperatures drop and food supplies fall to their lowest levels. Firewood is difficult to obtain and movement is highly restricted.

3.2 Ethnography

This thesis deals specifically with a rural Oromo community, which inhabits the rift valley area of Southern Ethiopia. The Oromo are one of the largest ethnic groups in Ethiopia (40% of the Ethiopian population). Geographically, they populate an area extending from the highlands of Ethiopia in the north, to Somalia in the east, to Sudan in the west and across the border to Lake Tana in Kenya to the south. Linguistically, Afan Oromo belongs to the Eastern Cushitic language groups, which include Somali, Afar and Rendille (Newman 1995).

Rural Arsi is populated by two groups of Oromo: the Arsi Oromo and Shoa Oromo. The Arsi Oromo are indigenous to the region; formerly living as nomadic pastoralists, they began to settle in low-lying areas during the mid-20^^ Century. In the 19^^ Century they were contacted by Muslim groups and adopted Islam (Trimingham 1965); however, today religious observances go little beyond fasting for Ramadan and avoidance of eating meat killed by a non-Muslim [Appendix I: Illustration 13].

The Shoa Oromo are migrants from the Eastern Shoa region who arrived in Arsi during the early part of 20^^ Century; they now engage in farming in the highland areas. Shoa Oromo practice Ethiopian Orthodox Christianity [Appendix I: Illustration 14]. The forms of Islam and Orthodox Christianity practised in the region retain many features of indigenous animist traditions (Levine 1965), e.g. female circumcision, bride abduction. Both groups populate the study villages (Table 3.1) and live in similar socio-economic circumstances. The main differences between the two groups of Oromo relate to cultural

31 Chapter 3: Study site rites of passage, e.g. circumcision and marriage practices (section 3.2.3).

3.2.1 Subsistence practices Traditionally, the Oromo were nomadic cattle herders, adopting a highly mobile life in scattered settlements throughout the highland area (Trimingham 1965). The districts of Hitosa and Dodota were inhabited only during the wet season; however, during the drier months the entire households would move with the cattle towards Lake Ziway in search of water. Due to the shortage of grazing land the Arsi Oromo began to settle and cultivate the dry savannah during the mid-20^*’ Century. Some may have adopted crop cultivating as early as the 1920-3Os, following the immigration of Shoa Oromo into Arsi highlands from the north (Selinus, Gobezie et al. 1971). However, following the political events of 1974 settlement became universal. During the 1970s and 1980’s the Marxist-Leninist Dergue regime introduced a number of land reforms, including villagisation, land re-distribution and the formation of Producer’s Co-operatives in rural communities across Ethiopia. Between 1974-1985 a significant number of these resettlement initiatives affected Arsi, forcing all Oromo into well-defined settlements with both agricultural and cattle breeding subsistence practices (Cohen and Isaksson 1987). Appendix II provides a summary of regional historical dates and events.

Today, the Arsi and Shoa Oromo represent a settled agro-pastoral population. The subsistence economy is predominantly based on crop production, with wheat, and barley being the main crop types in the highland areas, and maize is farmed in lowland areas. Cattle herding is still prevalent, playing a central role in Oromo cultural life; cattle are the primary unit of exchange at marriage (Baxter 1996). However, over the last decade cattle numbers have declined, as grazing land has become increasingly scarce (described in section 3.2.4).

3.2.2 Village/Household residence

In low-lying areas, the traditional Oromo settlement pattern has prevailed, consisting of scattered homesteads scattered over a large area forming a small hamlet. However, land reforms have resulted in some villages adopting more defined homesteads laid out in a grid pattern. Villages are large, usually comprising of between 800-1500 people. The

32 Chapter 3: Study site average homestead consists of one to four circular huts and is fenced by thorny acacia (Appendix I: Illustration 5). The roof of each hut is thatched and, along with the floor and walls, is sealed with dung. Each hut consists of one large communal room without windows, in which all activities take place (cooking, eating and sleeping). The furniture is limited to a small hearth, wooden stools and a few hides; only the wealthiest families own a bed or a table [detailed in section 3.1.4]. The total composition of each village is outlined in Table 2.1.

The household composition has the characteristics of an extended family structure, which is typical of a rural economy where the family is the unit of both production and consumption (Kebede 1978). The household generally consists of the nuclear family (husband, wives and children). The various other members of the household (e.g. in­ laws and co-wives) are likely to live within the homestead, but in separate huts.

The declining ratio of land to people has forced some men to migrate to seek employment in urban areas or to enlisted in the national army, from which many have failed to return. In the survey 9% of reproductive aged women stated that there was no male head of household. In these households, the responsibility for all subsistence and income-generating activities falls to women (Appendix I: Illustration 14). Any movement of people between villages occurs through marriage alliances; primarily through a system of exogamy. 96.4% of the male population were bom in their village of residence; however, only 13.8% of married females had remained in their natal village.

3.2.3 Marriage patterns and kinship structure

Among the Arsi Oromo, marriage is exogamous, inheritance patrilineal and residence patrilocal (Baxter 1996). Land is inherited through the male line, around 0.25 of a hectare is given to each son at marriage; however, the oldest son is likely to have the largest share. Each village consists of related kin groups in a five level organisation: the household/nuclear family (mana); the father’s and married son’s households (warra); a wider kin group including unmarried sons and maternal relatives (anaa); a minor lineage (balballa); and the highest level of kin group, the clan or gossa (Terefe 2000).

33 Chapter 3: Study site

Nuptiality has two major features: universality and early marriage (Hailemariam and Kloos 1993; Dagne 1994). At a national level as many of 10% of all 11-14 year old Ethiopian girls are already married and cohabiting (UNICEF 1993). In Arsi it is common for the bride to be in her teens and for the groom to be in his late twenties. Mean age at marriage for married women in the study population was 15.95 ±4.15 and for married men was 23.12 ±4.13. Figure 3.2 demonstrates the distribution of age at first marriage for the ever-married women (<55 years) enumerated. Over 90% of women were married before the age of 20.

Fig 3.2 Distribution of age at first marriage for all ever-married women (n=1566)

35

10 12 14 16 18 20 22 24 26 31

Age at first marriage

The two most common forms of marriage arrangement that exist in this region of Arsi are bridewealth and exchange marriages. Gabbara or bridewealth marriages are those in which marriage is arranged by purchase, with a payment made in cattle or money to the bride’s family. In 1999-2000, the size of the average bridewealth was 1000 Ethiopian birr (approximately 125 USD) and up to 10 cattle. Marriage occurs between families from different clans (the bride and groom have no common relative for seven generations through the patriline), often establishing affinal kinship links between villages. Prior to the water development, marriage alliances facilitated important rights

34 Chapter 3: Study site of access to additional water sources in other villages during the driest months. Post- marital residence is patrilocal [Appendix I: Illustration 11].

Walgara ‘I give you my child, now give me your child’ is an exchange marriage arrangement in which two families each exchange a son for a daughter and no commodities are exchanged. This form of marriage may occur between neighbouring houses in the same village, especially among poorer families who have no cattle available for bridewealth transactions. Post-marital residence is generally matrilocal.

For the Muslim (Arsi) Oromo marriage is indissoluble once the final sacrificial ceremony has been performed (Trimingham 1965); however, widow remarriage, e.g. levirate or dhaala (inheritance of dead brother’s wife), is common. Divorce may occur among the Shoa (Christian) Oromo. In the study villages only 8.4% of ever-married women reported that they had been married more than once. The family ideal is for a man to have several wives and many children; however, in practice polygyny can only be achieved by a rich minority. In this study around 25% of marriages reported were polygynous.

Pre-marital and extra-marital unions are strongly prohibited for women and social laws are strongly enforced. Female circumcision (usually a partial or complete cliterodectomy) is a universal requirement for both Arsi and Shoa Oromo (Terefe 2000). Among the Shoa Oromo, female circumcision occurs in childhood; however, Arsi Oromo women are not circumcised until a month before their marriage ceremony, at which point a girl’s virginity will be checked.

3.2.4 Socio-economic conditions Nationally, electricity, piped water and sanitation facilities are concentrated in urban households [Ethiopia], 2001 #632]. Rural housing characteristics are generally poor (92% of households have no toilet facility; <1% have electricity; 13% own a radio); however, the socio-economic conditions of the study site indicate that it likely to represent one of the poorer regions of the country. Few households are furnished with any of the household items, which have been used to categorise socio-economic status

35 Chapter 3: Study site in national surveys, e.g. radio, table, bicycle, latrine (Table 3.2). Households’ main economic activities relate to agricultural products; however, only 19.8% of households retain cattle, despite their social and economic importance. There remains little socio­ economic variation between households or villages. Following a programme of land re­ distribution in the 1980s (Appendix II) and in the absence of dramatic improvements to the agrarian economy, the degree of socio-economic differentiation between households is limited. Since villages are remote, villagers usually travel on foot (1-4 hours) weekly to attend agricultural markets in neighbouring towns. Few opportunities exist for pursuing any other income-generating activities, since villages remain without transport or communication links.

Table 3.2 outlines the socio-economic characteristics of the households included in the village survey (n= 1574). Villages are divided into those with, and those without, access to the new taps. While villages were selected for greatest comparability (Table 3.1) it is clear that some differences exist, e.g., villages with taps have a larger population of orthodox Christians, and more farmers dependent on wheat/barley cultivation. Overall, fewer households in the tap villages own livestock, however donkey ownership is more common (which may relate to water collection, see Table 3.5). Levels of household item ownership are comparable.

Table 3.2 Distribution of indices of household socio-economic status

Villages without taps (n=792) Villages with taps (n=782) All N % n %% Religion Orthodox Christian 82 10.4 380 48.6 29.4 Muslim 710 89.6 402 51.4 79.6 Economic activity Maize farmer 417 52.7 176 22.5 37.7 Wheat/ Barley farm. 373 47.1 599 76.6 61.8 Teacher/Govn.work 2 0.3 7 0.9 0.5 Livestock owner 728 91.8 636 81.3 86.8 Cattle None 118 14.9 194 24.8 19.8 1-5 500 63.1 527 67.4 65.2 6+ 174 22.0 61 7.8 14.9 Donkey None 217 27.4 355 45.4 36.3 1 397 50.1 315 40.3 45.2 2+ 178 22.5 112 14.3 18.5 Horse 38 4.7 13 1.7 3.2 Plough 703 88.8 744 95.1 91.9 Table 21 2.7 49 6.3 4.4

36 Chapter 3: Study site

Wooden/Metal bed 353 44.6 337 43.1 43.8 Radio 99 12.5 133 17 14.7 Bicycle 9 1.1 18 2.3 1.7

Use of pit latrine 0 - 69 4.4 4.4

3.2.5 Education levels The 1994 Ethiopian Population and Housing Census (CSA 1998) indicated that 3.4% of the entire school-aged population of Oromiya enrolled in high school (grades 9-12). In the study villages considerably fewer (only 2.4%) of the school-aged population reported ever attending high school. The majority of the adult population (>10 year olds) of these villages (66.7%) receive no schooling. Moreover, educational opportunities are generally biased towards males [Chapter 7]. Around 10% of females receive some form of formal education compared with 50% of males (Table 3.3). Primary level education is only available in market towns, and the nearest high school is over 30km away in the city.

Table 3.3 Level of educational attainment for adult population (greater than 10)

Level of education Males % (n=2999) Females % (n=2952)

None 51.0 87.7 Elementary (Grade 1-4) 30.8 13.8 Junior High School (5-8) 13.8 3.0 High School (9-12) 6.6 0.01

3.3 Health and population The Ethiopian population has been used in a number of studies to characterise a region with a high prevalence of moderate malnutrition. Prior to this study, no nutritional surveys had been undertaken in Hitosa and Dodota woredas. However, studies in the neighbouring region of Sidamo indicate that the southern parts of the country have a high prevalence of undemutrition (Ferro-Luzzi 1990; Ferro-Luzzi 1990; Branca 1993) and nutrient deficiencies, e.g. vitamin A deficiency (Desole, Belay et al. 1987). Vitamin A deficiency is associated with monograin subsistence (and lack of fruit and vegetables) and is known to aggravate diarrhoea, measles and acute respiratory infections, which are major causes of infant/child mortality, as well as causing xerophthalmia (blindness).

Although Arsi is not traditionally considered one of the most drought-prone areas of the

37 Chapter 3: Study site country (such as Welle and Tigray), there are a number of reasons why this region has been vulnerable to food insecurity since the 1990s. A declining ratio of land to people, lower land productivity, few relief and development NGOs, low investment in roads and other infrastructure are features prevailing in this region. Furthermore, increased outmigration to the capital, Addis Ababa, has eroded the natural resource base and traditional social mechanisms employed by people to cope with rain failure and poor harvests. A survey of relief food requirements in Arsi zones in 1999 reported increasing food insecurity and shortages in all woredas due to the anomalies in the rainy seasons (failure of the arfassa short rains in March) and a fungal disease (rust) in the wheat crop (Belay 1999). The assessment estimates that the number of people requiring relief assistance rose dramatically from 35,000 in 1998 to 150,0000 in 1999. Nutritional levels for women in the sample are described in Chapter 5.

3.3.1 Health services Health service coverage in Ethiopia is generally poor, with an urban-biased distribution. Adequate health provisions are available only to 45% of the population. Services targeted at women are less frequent; 25% of women receive antenatal care and <6% use family planning (CSA 1998). Following a programme of infrastructural development, including road construction schemes in Arsi in 1994, several new government clinics were established in the neighbouring market towns.

The villages in the Hitosa woreda are serviced by four health clinics in Iteya Town, Wolonkomi, Ligaba and Kulumsa; and those in the Dodota woreda attend clinics in the market town of Derra, Amudee and Doworro. The nearest hospital is in the zonal capital town, Asella, approximately 30km from the villages. Women interviewed expressed a reluctance to seek health treatment in these towns, since both the treatment and the transportion can be costly. Of those women who reported ill health (n= 449), over 60% had sought no treatment for their illness, 6.0% had treated themselves with traditional medicines and only 34% had attended a clinic. However, the severity of their illness had confined 62.6% of women to their bed and 34% had been ill for over a week.

38 Chapter 3: Study site

As yet no detailed health survey has been undertaken in the region, however, national surveys have highlighted diarrhoea, gastroenteritis and parasites as being the most serious health problems, especially for young children living in rural areas (Yohannes, Streatfield et al. 1992). Chapters 5 and 6 include sections describing the health characteristics of women and children respectively.

3.3.2 Population levels As with other developing countries, time series data on fertility in Ethiopia are scarce. However, the available studies from Ethiopia give a general impression that fertility is high and has been rising (Mammo and Morgan 1986). Ethiopia (excluding Addis Ababa) has the second highest fertility rate in Africa (TFR = 7.1) and population growth rates are among of the highest in the world [2.9% in 1994] (CSA 2001). Maintaining a similar rate, the current population, of around 70 million, is likely to double in less than 25 years.

National concerns about growth rates are only very recent and the ability/desire to reduce number of births is as yet limited to small socially mobile minorities (Bantje 1995); only 8% of the total population use some form of family planning (CSA 2001). The low rates of contraceptive prevalence in rural areas reflect social prestige and economic security associated with large numbers of children and a negative attitude toward limiting family size. Less than 1% of Oromo women in the present survey had ever used any form of contraception, not only because short birth-spacing and large family size is desirable, but also due to the lack of family planning services available to them.

Rapid population growth in this region has been attributed to high and increasing fertility rates and declining mortality rates among the rural population (Hailemariam 1991). In the national DHS Survey, the region of Oromiya recorded the highest fertility in the country (TFR = 6.4) (CSA 2001). The fertility and mortality levels at the study site are discussed in Chapters 4 and 5 respectively.

39 Chapter 3: Study site

3.4 Problems of water supply The communities of the Hitosa and Dodota district suffer from acute and regular water shortages. These water problems date from the 1930s when an Italian entrepreneur and Ethiopian landowner went into partnership to exploit the farming potential of the area. To obtain the workers for their venture they moved people from where they were living near rivers and springs and settled them on lower, more fertile land (Silkin 1998). While the region receives good rains for farming, the main rivers that are its source of domestic water flow around the district rather than through it, and consequently distances covered to collect water are considerable. The closest perennial rivers to the district are the Gonde River which flows about 30 km south-west of its boundary, the Awash River is 50 km north-east and the Wodecha River about 40 km south-east.

Trips to collect water are infrequent, since the journey is long and arduous. During the driest seasons, women collect an average of 40 litres/day for domestic use, which supports around 7 family members. This should permit 5.7 litres daily per capita usage; which is considerably lower than the national average (8-9 litres) (Teka 1993) and the recommended target advocated by non-governmental organisations (20-40 litres per capita). In the absence of water tap stands, water rationing continues to be a common feature of household life.

3.4.1 Water collection Across Africa rural women have been traditionally responsible for all household tasks, including water collection, for which they have received little human or technical assistance. The disproportionate amount of time that women spend in water collection has been well documented across Sub-Saharan Africa (Chisholm 1968; White, Bradley et al. 1972; Feachem, Bums et al. 1978), including Ethiopia (Kebede 1978). Women continue to be the primary drawers of water, a role often considered an exclusively female one.

In an extensive study of water use in East Africa, White et al. (White, Bradley et al. 1972) demonstrated that water carriers in some rural communities spend up to 4.5 hours per day collecting water. In Hitosa and Dodota the acute water shortage means that the majority of women are required to walk a minimum of 3 hours a day on a trip to obtain

40 Chapter 3: Study site water from river or spring sources (Table 3.1). During the wet season local ponds are dug and utilised; however, these dry up later in the year [Appendix I: Illustrations 6 & 7]. During the dry season some women report spending up to 8 hours/day collecting water, from spring sources up to 30 km away.

The loads that women bear in numerous household chores are often considerable (White, Bradley et al. 1972). Female fuelwood carriers in Addis Ababa are known to transport loads heavier than their own body weight (Haile 1989). In Arsi, women transport water in containers strapped to their backs, either in a plastic jerry-can, holding 20-25 litres, or a traditional clay water pot {insera), containing 15-20 litres [Appendix I: Illustrations 8 & 9]. The average water-carrying load for women is around 40% of their body weight (Table 3.4).

Women are required to make at least one trip per day to collect household water; and over a third receive no assistance (Table 3.5). Women from the wealthiest households may use donkeys to transport the loads; however, over 10% always carry the loads themselves. Daughters and young children or other kin may sometimes accompany women, however they rarely assist with load carrying.

Table 3.4 Water carrying workload for women in the dry season months (bonaa)

Sample Mean Age Mean Average water load Load % o f Num ber of Weight (kg) carried (kg per trip) body mass weekly trips

464 32 ±9.98 49.18 ±5.94 20 40.67 8.98 ±4.21

Table 3.5 describes the methods of water carrying employed by women at the survey date, during the dry season months. Villages are divided into those with, and those without, access to the new taps. In villages using traditional water sources, fewer women use the heavy clayinsera water-carriers and more use donkeys to assist them with load carrying. Moreover, their husbands are more likely to help in water collection duties. Access to taps is associated with women undertaking on average one extra trip to collect water a week. However, this is likely to contribute only 40 litres extra water for the entire household per week.

In certain populations women may have found energetically economic ways to carry

41 Chapter 3: Study site loads (Maloiy, Heglund et al. 1986; Jones 1989); however, for the most part the energy expended and activity patterns undertaken by subsistence-farming rural women can be compared to those of endurance sportswomen (Rosetta 1993). Furthermore, by expressing the time and energy spent on water collection as a percentage of total available daytime time and energy, individual rural households can expend as much as 14% of total energy on water collection (White, Bradley et al. 1972). The time and energy costs imposed on women water-carriers have been associated with a number of negative health, economic and social outcomes (see Curtis 1986 and Lukmanji 1992 for reviews).

Table 3.5 Description of water-carrying methods in both villages with and without taps

Viilages with taps (n=599) Viiiages without taps (n=967)

Collection 88.5% Jerry Cans 94.9% Jerry Cans method 11.5% Insera 5.1% Insera ^ Litres 37.39 ±11.27 37.77 ±8.94 collected /trip t=-0.744 p=0.457 Weekly trips 9.75 ±4.54 8.05 ±3.60 t=-8.194 p=0.000** 4/ Donkey use 75% always use 91.8% always use 15.2% never use 5.2% never use 4/ Assistance 29.9% have no 2.3% co-wives 31.0% have no 2.1% co-wives assistance 27.0% daughters assistance 26.2% daughters 15.9% sons 12.6% sons 10.5% husbands 18.0% husbands^ 11.7% servants 8.3% servants Notes: 4» significantly lower in villages without taps (p=<0.05); t significantly higher in villages without taps (p=<0.05)

3.5 Development infrastructure

The widespread media attention surrounding the droughts of the 1980s, and 30 years of civil conflicts, has resulted in the image of Ethiopia becoming synonymous with poverty, famine and political disturbance (Appendix II). While other rural areas of East Africa have benefited from overseas development aid, the civil instability pervading in the Horn of Africa since the 1970s has restricted the amount of humanitarian assistance reaching rural Ethiopia. However, since the overthrow of the communist regime and the instatement of a democratic government in 1991, overseas development agencies have begun to fund development initiatives across all regions of the country. WaterAid, a UK-based international water development organisation, has been one of the first

42 Chapter 3: Study site international donor agencies to provide financial and technical support for a number of water supply projects in Ethiopia.

3.5.1 The Hitosa gravity water supply project

The international development charity, WaterAid, and their Ethiopian partner organisation. Water Action, have been funding and managing a water supply project in the Hitosa and Dodota woredas of Arsi Zone, Oromiya since 1993. The function of the project has been to provide a domestic water supply for 28 villages, by installing 122 public tap stands in villages [Appendix I: Illustration 10]. Taps are part of a gravity water supply scheme, in which a single pipeline is laid through a number of villages through a descending gradient and the taps are activated simultaneously. The primary objective was to reduce the physical stress on women and the time spend in collection, by making clean water available at reasonable distances (100-500 m). Prior to the water development scheme, the journey time to collect water from river sources could be as long as 4 hours for a round trip. Since the implementation of the Hitosa scheme, the journey times have been reduced to less than one hour per round trip (Silkin 1998). This project was designed as a single intervention, with no community-based activities that might alter other social services, such as providing healthcare or education. Significant migration to take advantage of the new water supply scheme is also not likely due to the acute shortage of land in the area, which prevents new settlement.

During Phase I of the project, the pipeline was laid and taps were installed and began issuing water in the villages in March/April, 1996 {Miyaza, 1988 in the Ethiopian Calendar). Phase II of the project commenced in 1999 and has aimed to establish a health and hygiene education programme in each of the villages; this includes providing materials for the construction of pit latrines. However, at the time of the survey only 4.4% of the households had access to toilet facilities and the programme of hygiene and health education had not been fully implemented. Work to extend this pipeline commenced during 1999, in an adjacent district, between the towns of Iteya and Gonde, where villages continued to use traditional water sources.

43 4

The impact of labour-saving technology on

fertility differentials*

‘Women give birth to many children despite the hardship of our conditions..it is very confusing for us, because previously conditions were good, but women had fewer c h ild re n ’ [Male informant, Debula Saapo]

4.1 Fertility levels Micro-demographic studies undertaken by anthropologists provide useful data for examining the changing nature of rural communities, including adaptation to ecological, economic and social conditions (Howell 1979; Pennington and Harpending 1993; Clark, Colson et al. 1995; Hill and Hurtado 1996). This chapter includes sections which describe period fertility at Hitosa and Dodota villages at the survey date, and explore fluctations in fertility over time using birth history data. The main analyses investigate the impact of the new water supply scheme on fertility differentials, specifically birth- spacing, using multivariate event history techniques.

4.1.2 Period fertility rates

Fertility is defined by demographers as the production of a live birth, that is, a child bom alive (Pressât 1985). The total fertility rate (TFR) is frequently used as a measure of fertility for a population. It represents the number of children a woman could have during her lifetime if current age-specific fertility rates were to remain constant (Wood 1994). For the purposes of this study, the total marital fertility rate (TMFR) is calculated, since the taboo against pre-marital sex is strongly enforced and marriage is universal [Chapter 3: Section 3.2.3]. TMFR is calculated as the sum of the age-specific marital fertility rates over the whole range of reproductive ages for one year.

* Analyses published in: Gibson, M & Mace, R. 2001 The impact of labor-saving technology on first birth interval lengths in Rural Ethiopia. Human Biology, v. 74, n. 1, pp 111-128. Gibson, M. A. & Mace, R. 2002. Labor-saving technology and fertility increase in Rural Africa. Current Anthropology, Aug-Oct. Chapter 4: Fertility

Age specific marital fertility (ASMFR) refers to the proportion of married women in a 5-year age group who have given birth in the year preceding the interview.

ASMFR = Number of births to married women aged x

Number of married women aged x at mid year

The entire data set of live births obtained from all reproductive-aged (<50) ever-married women is used to calculate age-specific fertility rates (n=1566). The TMFR for the study population is 7.86 and an inter-birth interval length of 30.55 ±9.29, indicating a high level of fertility relative to other recent regional surveys (Table 4.1).

Table 4,1 Data on fertility in Southern Ethiopia

Site Date Age at TFR TMFR le r Reference marriage Hitosa/Dodota 1999/00 15.85 7.86 30.55 ±2.46 ±9.29 Arsi 1991 7.21 Hailemariam,91 Oromiya 1995-00 16.4 6.4 31.4 CSA (DHS), 01 Ethiopia 1995-00 16.4 5.9 33.6 CSA(DHS), 01 months since the preceding pregnancy that

4.1.2 Fluctuations in fertility rates over time To examine the changing fertility patterns during the 20 years preceding the survey, total marital fertility is calculated as the proportion of women giving birth, classified by two-year groupings. Figure 4.1 illustrates the fluctuations in fertility levels across this time period. The main trend is towards declining fertility during the 1980s, followed by an incline during the mid-1990s, coinciding with the introduction of new development technologies.

Fig 4.1 Total 9 marital fertility rates (1980- 8.5 1999)

8 q: u_ famine s 7.5

water project ^ 7

6.5 81-82 83-84 85-86 87-88 89-90 91-92 93-94 95-96 97-98 n (births) = 446 506 507 571 562 630 642 699 780

45 Chapter 4: Fertility

4.2 Does labour-saving technology impact on fertility? Across the developing world, labour-saving technologies have been designed and implemented to introduce savings in the time and energy that women allocate to work. In rural Arsi, southern Ethiopia, a recent water supply scheme has reduced long arduous trips to obtain water and is associated with considerable improvements in women’s energy budgets. Assuming that the time and energy saved is not diverted to other energetically costly activities and nutritional levels remain constant, evolutionary life- history theory predicts that this energy may be diverted into reproductive effort.

Field studies in human reproductive ecology have revealed that fecundity is responsive to changes in maternal condition (Little, Leslie et al. 1992; Hill and Hurtado 1996). Other clinical trials have identified the physiological pathways along which energetic factors influence reproductive function [described in Chapter 1]. A positive energy balance, attributed to the combined effect of improved seasonal workloads and nutrition, is associated with higher fecundity (Ellison, Peacock et al. 1989; Bailey, Jenike et al. 1992; Panter-Brick, Lotstein et al. 1993). Although the effects of short-term seasonal alteration in energy levels have been explored, the effect of long-term changes under conditions of poor nutrition is less well understood. No study has attempted to identify whether physiological changes associated with workload affect fertility at a population level - that is, translated into a measure of reproductive outcome, the number of live births.

The aim of the following sections of the chapter is to detect any effects of the installation of village water taps on birth spacing, both through the timing of the first birth and intervals between subsequent births. The other bio-behavioural changes associated with water development, which also influence fertility [outlined in Chapter 1], will also be explored e.g. marriage patterns, breast-feeding practices and mortality rates.

A new labour-saving grain mill had no impact on Mayan birth interval lengths (Kramer and McMillan 1998; Kramer and McMillan 1999). However, due to constraints of the data available to them (the grain mill had been installed decades before their survey), they were unable to control for secular trends in fertility. In this thesis the water scheme is a recent event and it is possible to pinpoint the month of arrival of the water taps

46 Chapter 4: Fertility

(March 1996). By entering this event as a time-varying covariate in an event history analysis, demographic change is sought at that exact point for each woman. Furthermore, since some villages continue to use traditional water sources, these serve as a proxy for conditions prior to development.

Using the event history calendar data collected in the demographic survey [Chapter 2, Appendix VII], hazard regression techniques (event history analysis) are employed to assess the impact of the installation of tap stands on two correlates of recent fertility: the timing of the first birth, and the length of subsequent inter-birth intervals.

4.3 Main analytical methods Using univariate life table techniques, which incorporate both uncensored and censored lengths of birth interval, the median length of first birth interval by different characteristics (estimated from the survival function) are calculated to provide a crude indication of the variability in birth interval length (SPSS 9.0). However, by using additional multivariate hazards models, it is possible to assess the partial effects of several factors, including time dependent ones, on the length of the birth interval (Allison 1984).

Discrete-time methods of event history analysis, using logistic regression to estimate the multivariate model, are employed in the main analyses (PROG LOGISTIC in SAS v.6.12). These directly assess the effects of the improved water supplies on: (1) the timing of the first birth; (2) the length of open birth interval; and (3) the duration of post-partum amenorrhoea, after controlling for other factors.

The models describe the observed and multivariate results of the analyses in the form of exponentiated coefficients for the log odds. Coefficients are converted into an odds ratio relating the presented effect to the omitted category by calculating Exp(P) for each coefficient in the original equation (dummy variable logistic regression).

4.3.1 Data set For the purposes of the analysis of the timing of the first birth, the data set is limited to a sub-sample of 366 women who experienced a first marriage during the 6-year observation period (Table 4.3). This includes both, those women who experienced at

47 Chapter 4: Fertility

least one birth and those cases in which a woman had not experienced a first birth at the time of interview (right-censored). Married women who failed to give birth within 4 years of marriage are excluded. Women known to be practising any method of family planning at the time of interview are included, based on the assumption that family planning is unlikely to be used before the birth of the first child. Overall, contraception prevalence is extremely low in this population (<1% of women reported ever having used contraception). A total of 107 of the women had also been measured for height and weight during the anthropometric survey; these are included in the analysis of anthropometric variables (Table 4.4).

For the analyses of inter-birth interval lengths, the data set is limited to non-pregnant women who experienced at least one birth within the 60 months prior to the survey date (Table 4.7). All women known to ever have used family planning are excluded. In this case, the final sample for analysis is restricted to 893 women. A total of 324 of these women had also been measured in the anthropometric survey and therefore could be included in the analysis of anthropometric covariates (Table 4.8).

4.3.2 Dependent variables The month of birth of the first child is the dependent variable for the analysis of the timing of the first birth (or first birth interval length). The first birth interval is the time elapsed between marriage and first birth. It is of interest to demographers since it is an important predictor of lifetime reproductive outcome, affecting a woman’s subsequent birth-spacing and child-bearing pattern (Bumpass, Rindfuss et al. 1978; Trussed and Menken 1978).

The open birth interval (the time between the last birth and the survey date) is used as a correlate of recent inter-birth-spacing. Analysing the open birth interval is a useful way to establish temporal relationships between the on-going process of fertility and women’s current status (Srinivasan 1972; Feeney and Ross 1984). Furthermore, it includes a large sample for analysis. In this analysis the dependent variable is the time elapsed since last birth - that is, the hazards analysis is modelling the monthly risk of having no birth from the last birth to the survey date.

48 Chapter 4: Fertility

Finally, the timing of the return to menses (post-partum amenorrhoea) following each woman’s most recent birth is also modelled, since it is a major component of the period of post-partum infertility. The timing of return to menses has previously been linked to infant feeding patterns (Jones and Palloni 1989) and to child death and maternal nutrition (Huffman 1987). In this case, the dependent variable is the month of resumption of menses following the last birth.

4.4 Independent variables The independent variables are broadly categorised as socio-ecological, demographic and energetic.

4.4.1 Socio-ecological covariates Rural Arsi is populated predominantly by two groups of Oromo agro-pastoralists: the indigenous Arsi Oromo (Muslim) and the Shoa Oromo (Orthodox Christian) [outlined in Chapter 3: Section 3.2]. The variable for religion is entered into the analysis as a proxy for ethnic grouping, since ethnicity and religious beliefs are analogous. However, this covariate is removed from the current analysis of first birth interval lengths since it consistently had no significant effect on first birth interval length in previous models.

The seven villages included in the study are located in two ecological zones (Table 3.1). The lowland areas are subject to irregular rainfall and soil erosion; in these areas maize is the predominant crop type. The highland areas experience more regular rainfall and support a subsistence economy based on a range of crop types including wheat, barley, and teff. For the purposes of the analyses, village characteristics are entered into each model, either by individual village, to control for unobserved heterogeneity between villages, or categorised as a dichotomous covariate for village altitude.

Without improved access to the local market economy, the degree of socio-economic differentiation within villages remains limited. The categorisation of socio-economic status, used in other studies, based on household items and dwelling characteristics are inappropriate in this context. In this case, household herd size is used as a proxy for social status, since cattle ownership is associated with inherited wealth and social prestige, and is categorised as three groupings: low status (= no cattle), medium (= 1-5 cattle) and high (= 6+ cattle). An additional variable, radio ownership, is included as a

49 Chapter 4: Fertility further indicator of household economic status. These current status variables are assumed to be invariant over the 6-year observation period based on the assumption that ^vithin a rural subsistence based society, the economic condition of any household is unlikely to fluctuate drastically over such a short period of time.

A further social covariate describing level of maternal education was included in earlier models (not presented here) but is excluded from current analysis, since it consistently had no significant effect on any measures of fertility. Previous studies among developing world subsistence populations have indicated that female educational attainment is correlated with birth spacing, with high levels of schooling being associated with wider birth-spacing practices (Fricke and Teachman 1993; Nath, Land et al. 1999). This effect may not be evident in the study population since too few Oromo women surveyed received any form of schooling [Chapter 3: Table 3.3].

4.4.2 Demographic covariates To control for any historical secular trend in the data spanning 6 years a covariate for marriage cohort is included in the first birth analysis. Marriage cohort is categorised according to year of marriage (1) 1993-94 (2) 1995 (3) 1996 (4) 1997 (5) 1998 (6) 1999-2000. To investigate the partial effect of maternal age, women are classified into four age groups at the start of the interval. For the analysis of the first birth, mother’s age refers to her age at marriage (1) <= 14 years (2) 15-16 years (3) 17-18 years (4) >= 19 years. Since maternal age and parity can affect the length of the birth interval, in the analyses of inter-birth intervals mothers are classified both in parity groups, according to number of live births, and into age groups, by decade of life.

A binary covariate to describe marriage season is included in the analysis to control for any effects of seasonality. Although women do not work heavily in the fields, seasonal changes in subsistence ecology relating to other workloads, food availability and disease may impact on fertility (Leslie and Fry 1989; Hurtado and Hill 1990; Bailey, Jenike et al. 1992; Panter-Brick, Lotstein et al. 1993). The dry season is defined as the post harvest season (October-April). The wet season extends from May, through the wettest summer months, to September [Chapter 3: Section 3.2].

50 Chapter 4: Fertility

In the analyses of the correlates of inter-birth interval length, a dichotomous covariate for polygyny, determined by the current marital status of each woman’s husband (at the date of survey/ end of the open interval), is also included. The death of an infant may also strongly influence the length of the birth interval (Koenig, Phillips et al. 1990). The survival status o f the index child, whose birth opens the interval, is coded as a time- varying covariate.

4.4.3 Energetic covariates The use of time-dependent covariates, namely the timing of the water tap installation and function, is an important feature of this present analysis. Using event history analysis it is possible to examine whether the change in the state of the covariate (function of the tap) influences the hazard rate/risk of experiencing a dependent event (birth) [described in Chapter 2]. Function of water taps is entered into the analysis as a time-varying covariate to assess the monthly risk of experiencing a birth/retum to menses since the water taps began operating in March 1996. A binary covariate water access is created for each woman; each person-month without a birth is coded as either ‘9’ if it occurs before (or < 9 months after) and V ’ if it occurs 9 months after water taps were installed in that woman’s village. In the four villages excluded from the water supply scheme, this variable is coded ‘0’ for the entire observation period.

Ten percent of women collect and carry water on their backs without assistance from either kin or donkeys [Chapter 3: Table 3.5]. These women carry loads that are on average 40% of their own body weight on return journeys taking up to 1-3 hours. If workload affects birth spacing, then the method of water collection is likely to explain some of the variation between women. The dichotomous variable water carrier is created as a proxy for the women’s water-carrying workload. Each woman is classified according to whether she exclusively carried water on her back, or whether kin or donkeys assisted with load carrying.

To examine the separate effect of women’s nutritional status on reproductive function, the variable body mass index, (5M/=weight in kg/height in metres^) is calculated for the sub-sample of women included in the anthropometric survey (n= 324). Body mass indices are categorised into three groups on the basis of the recommendations of (Ferro-

51 Chapter 4: Fertility

Luzzi, Setter et al. 1992): chronic energy deficiency (< 18.5), moderate (18.5-22), and highest levels of nutrition (> 22). An additional measure of stored energy is obtained from skinfold measures, indicating the thickness of adipose tissues, is included in the inter-birth interval analysis. Skinfolds are calculated as the sum of triceps and subscapular skinfolds: low (< 20 mm), and high (>= 20 mm). Although the nutritional status of each woman is unlikely to have remained constant over the entire observation period (e.g. fluctuating between seasons), this current status measure, broadly categorised, may be used as an overall indicator of variation between women.

Categorised length of previous birth interval is also included in the inter-birth interval analysis as an indicator of a woman’s recent birth spacing pattern. This variable is available only for a sub-sample of women who had experienced at least two live births (n= 684). In addition, controls for time and time squared (in months) are entered into all the models for analyses, since the risk of returning to menses or giving birth is likely to vary as a linear function of the length of exposure.

4.5 Timing of the first birth

‘A woman m ust marry when her breasts stand firm It Is better to marry women as early as possible to avoid pre-marltal sex. If the bride Is not a virgin then she brings shame to her family and all the presents given by the groom should be returned’ [Male informant, DS]

The birth of the first child is an important life event for a recently married woman, since it assures her position within her new household. A woman who fails to give birth to a child within four years of marriage is at risk of suffering the consequences of being labelled 'mertifltuu' (infertile). For Arsi Oromo, this may result in her husband taking an additional wife; while, for Shoa Oromo, she may be divorced and sent back to her parents.

Previous demographic studies have contributed to the understanding of first birth intervals through the analysis of socio-demographic variables, for example, age at marriage, maternal age, education and living arrangements (Fricke and Teachman 1993; Nath, Singh et al. 1993; Nath, Land et al. 1999). However, none to date have examined the effects of maternal energetic stresses on the timing of the first birth.

52 Chapter 4: Fertility

4.5.1 Age at marriage Behavioural changes associated with the new water development scheme may also alter fertility, such as marriage patterns and age at first intercourse. Access to labour-saving technology in a rural Mayan village was associated with an initial drop in age at first birth following changes in decision-making about when young girls leave home (Kramer and McMillan 1998; Kramer and McMillan 1999). New educational opportunities associated with the improved communication links with neighbouring market towns may contribute to a change in marriage patterns for the Oromo. Figure 4.2 illustrates the steady secular trend towards increasingly age at marriage for women over the 30 years prior to the survey; this has included a steep incline during the 1990s.

17.5

Fig. 4.2 Distribution of female mean age at 17.0 . marriage 1964-2000 (n=1566)

15.5

15.0

14.5 66 7363 78 83 88 93 98

Year of marriage

4.6 Analyses of the first birth interval 4.6.1 Univariate analyses Comparing the demographic characteristics observed in villages with and without access to taps can serve as a rough proxy for conditions before and after the installation of taps. Table 4.2 presents mean values for age at marriage and median values estimated from the survival function for age at first birth, and length of first birth interval between villages according to water accessibility across the observation period. Values are calculated for cohorts of women first married before and those married after March 1996, the date of water point installation. The results suggest that several demographic changes have occurred within those villages benefiting from improved water supplies. Since water taps were installed, mean age at first marriage and subsequently mean age

53 Chapter 4: Fertility at first birth have increased. There is some indication that length of first birth intervals may have shortened; however, the size of this effect increases once one lowland village suffering recent crop failures (Debula Saapo) is removed from the sample. In those villages continuing to use traditional water supplies across the entire observation period, there is no indication of change in age at marriage, age at first birth or shortened birth intervals.

An association between first birth interval length and tap installation does not necessarily indicate that the decline in birth interval length is causally related to an improvement in women’s workloads, but may be an effect of later age at marriage or change in other covariates. Adolescent subfecundity may be present in this population, since birth interval length varies by age at marriage and young women married <15 years experience an unusually long first birth interval. A period of subfecundability may last for up to two years following menarche (vom Saal and Finch 1994). To explore the partial effects of several covariates, multiple regression techniques are employed.

Table 4.2 Variation in mean age at marriage, median age at first birth and length of first birth interval within and between villages since tap installation (March/April, 1996).

Date Village Access to Water Without taps With taps (all) With water taps (exci. Debula Saapo)

Mean age at Before 1996 16.45 n=100 15.93 n=74 15.84 n=60 m arriage ±1.93 ±1.79 ±1.75 After 1996 16.68 n=95 16.92 n=97 16.92 n=64 ±2.03 ±2.09 ±2.09

Median age Before 1996 17.60 n=99 17.23 n=73 17.16 n=59 at first birth After 1996 17.68 n=57 17.97 n=71 17.90 n=46

Median Before 1996 15.60 n=100 14.83 n=74 14.75 n=60 length o f first After 1996 14.18 n=95 14.48 n=97 12.74 n=64 birth interval Note: * * Pre - Post 1996 Mann Whitney U test (p=<0.05)

The general characteristics of the median first birth intervals and the percentage of women who failed to give birth within the first 48 months of their marriage for groups of covariates included in the final multivariate analysis are presented in Table 4.3. Calculated using life table techniques, the median length of the first interval for the

54 Chapter 4: Fertility entire sample is 14.8 months, while only 2.64% of the women failed to give birth withi 48 months of their marriage.

Table 4.3 Univariate median survival time of first birth interval for covariates

Covariates Median n % with no births within 48 mths or up to censoring Overall 14.80 366 2.64 Marriage cohort 1993/94 15.50 58 0.00 1995 16.33 54 3.13 1996-water 14.00 66 6.06 1997 14.50 53 0.00 1998 13.75 55 11.81 1999/2000 15.34 70 10.74 Age at marriage <15 18.90 37 7.58 15-16 15.49 175 3.54 17-18 13.46 105 0.00 >=19 13.81 49 0.00 Marriage season Dry (Oct-Apr) 15.64 222 3.79 Wet (May-Sep) 14.10 144 1.04 Viliage Lowland 17.20 152 4.56 Highland 13.85 214 1.42 Herd size No cattle 15.73 164 2.40 1-5 cattle 14.22 162 1.10 6+ cattle 17.19 40 8.61 Water carrier On back 14.92 35 6.22 With help 14.79 331 2.06 Body Mass Index <18.5 13.33 17 0.00 18.5-22 14.46 71 1.53 >22 17.25 19 8.40

4.6.2 Multivariate analyses For statistical analyses to assess the effects of water point access on length of first birth interval after controlling for other factors, hazard regression models were estimated using two different specifications. Model 1 (Table 4.4) contains the results of the analysis using a discrete time logistic regression model of the effects of socio­ demographic variables, water point installation as a time-varying covariate and a dichotomous covariate to define method of water collection (n= 366). Model 2 (Table 4.5) introduces a covariate to control for women’s current nutritional status, body mass index for a reduced sample of women included in Model l(n= 107).

55 Chapter 4: Fertility

To test the assumption of proportional hazards, a separate model was run (not shown here) to explore any interaction effects between time (length of exposure since marriage) and the water point installation. There was no significant level of interaction, indicating that the assumption of proportional hazards is valid. Similarly, there were found to be no significant interactions between current nutritional status and time or between any other independent variables. However, an odds ratio greater than unity for time (months) and less than unity for time squared in all models indicates that risk of first birth has a curvilinear relationship with time (months since marriage), as identified in other hazards analyses of birth interval lengths (Wood 1994).

There is some evidence to suggest that there has been a decline in first birth interval length across all villages over the study period (since 1993) (Table 4.2). However, the results from the hazards models suggest that only women married the year before water point installation have a significantly altered risk of birth. These women have a lower risk of experiencing a birth in the months that followed marriage than the reference group (married in 1996, the year of tap installation).

In both the models, observed relationships suggest that women who marry at a greater age (>17) may have a higher probability of childbirth than those married at younger ages. Women married at <15 years of age are about one-third less likely than women in the following age group (15-16) to experience a birth per month following marriage; however this effect does not reach statistical significance (p=<0.05). Figure 4.3 illustrates the effect of age at marriage on mean completed first birth interval length. After including the nutritional status covariates in Model 2, the effect of maternal age at marriage becomes insignificant, which may be an artefact of the reduced sample size or implies that the effect of age on birth interval length is mediated through women’s nutritional status. However, women may have been measured at differing stages of adolescent growth and therefore age and energetic status variables may be interacting.

Season of marriage is also associated with probability of childbirth; women married during the wet season (May-Sept) are at a higher risk of birth per month than women married during the dry season. This season is associated with low water-carrying workloads as local groundwater is collected and improved nutrition as milk products are

56 Chapter 4: Fertility

readily available. This effect of marital season disappears in Model 2 (Table 4.5), which may be related to the reduced sample size or may suggest that the effect of season on first birth interval length is mediated through nutritional status.

Fig 4.3; Mean length of completed first birth interval by age at marriage

i/t .c c o E O) oc Ï sc t !5

15-16 17-18

Age at marriage (years)

In Model 1 (Table 4.4) the effect of the covariate for village on first birth interval confirms the expectation that ecological conditions associated with altitude would have an effect on fertility differentials. The observed effect reveals that women living in the three lowland villages are almost 50% less likely to experience a birth per month. It is these villages (particularly Debula Saapo) which have suffered the chronic effects of repeated crop failure in recent years. The higher monthly probability of birth in highland villages is illustrated in a Kaplan Meier plot in Figure 4.4. One effect remains consistently significant across models - that is, for women in Debula Saapo to experience longer birth intervals.

57 Chapter 4: Fertility

Fig 4.4 Kaplan-Meier plot of the effect of village altitude on probability of first birth.

c > D) Village altitude

■V Highland ; ' c

0.0 o Q. Lowland 2 CL 0 10 20 30 40 50 60

Months since marriage

Relationships between the first birth interval and household herd size are not observed in Model 1 (Table 4.4). However, in Model 2 (Table 4.5) there is a slight trend towards a lower risk of birth per month associated with cattle ownership, but sample sizes are smaller in this model.

Across models there is a tendency for women who exclusively carry water on their backs to be at a lower risk of first birth in each interval than women who are assisted by kin and/or donkeys. Women water carriers are on average 50% less likely to experience a first birth in each month following marriage; however, this effect does not reach statistical significance (p= <0.05). Water-carrying is likely to be a reliable indicator of energetic workload and not socio-economic status of the household, since the effect is in the opposing direction to that for herd-size, another measure of socio-economic levels (low herd size is associated with higher risks of birth).

There is a positive effect of water point installation on monthly risk of childbirth. The results suggest that improved access to water at any point since marriage almost doubles women’s risk of having a first birth per month. A Kaplan Meier plot (Figure 4.5) clearly

58 Chapter 4: Fertility illustrates this increased risk following improved water supplies, recording monthly survival functions for two marital cohorts, grouped by marriage date before or after water tap installation (excluding village Debula Saapo). In Model 2 (Table 4.4) the inclusion of nutritional status into the analysis reveals that body mass index has no significant affect on the first birth hazard, while the effect of the water point remains significant.

Fig 4.5 Kaplan-Meier plot of the monthly probability of first birth for two marital cohorts within villages with access to taps: women married before and after water tap installation in 1996.

Marriage cohort ■Ç .a married after c water D)>

o c o .2 married before o water OQ. Ql 0.0 0 10 20 30 40 50

Months since marriage

59 Chapter 4: Fertility

Table 4.4 Multivariate hazard regression model for first birth interval (Model 1: n= 366)

Variable Odds ratio P Marriage cohort: 1993-94 0.991 0.966 1995 0.647 0.046** 1996 1.000 - 1997 0.708 0.096t 1998 0.872 0.517 1999-2000 1.105 0.739

Age at marriage: <15 0.657 0.081t 15-16 1.000 - 17-18 1.425 0.021** >=19 1.145 0.511

Marriage season:

Dry 1.000 - W et 1.403 0.014**

Villages: D.Debeso (highland) 0.734 0.346 D.Gabrel (high) 0.803 0.497

T. Moyee (high) 1.000 - Hurturbe (high) 1.414 0.177 R.Michiko (lowland) 0.645 0.069t B.Washo (low) 0.554 0.021** D.Saapo (low) 0.368 0.004***

Herd size: None 1.000 - 1-5 0.922 0.601 6+ 0.693 0.122

Water carrier: No 1.000 - Yes 0.723 0.097f

Water point installation: Without access 1.000 - With taps 1.838 0.031**

Time since marriage 1.432 0.000*** Tim ef 0.993 0.000***

Intercept (Coefficient) -6.161 0.000***

Total cases (months) 6236 Births (n) 297

Note: Reference category has odds ratio of 1.000 tp=<0.1, **p=<0.05, ***p=<0.005

60 Chapter 4: Fertility

Table 4.5 Multivariate hazard regression model for first birth interval with anthropometric covariates {Model 2: n= 107)

Variable Odds ratio P Marriage cohort: 1993-94 0.780 0.583 1995 0.289 0.007*** 1996 1.000 - 1997 0.940 0.892 1998 0.771 0.507 1999-2000 0.669 0.545

Age at marriage: <15 0.674 0.415 15-16 1.000 - 17-18 0.722 0.369 >=19 1.674 0.169

Marriage season:

Dry 1.000 - W et 0.998 0.993

Viiiages: Daya Debeso (highland) 0.264 0.048** Daya Gabrel (high) 0.204 0.029**

Terro Moyee (high) 1.000 - Hurturbe (lowland) 1.533 0.391 Reissa Michiko (low) 0.825 0.713 Debula Saapo (low) 0.138 0.009***

Herd size: None 1.000 - 1-5 0.483 0.027** 6+ 0.403 0.0991

Water carrier: No 1.000 - Yes 0.508 0 .0 9 8 t

Water point instaiiation: Without access 1.000 - With taps 3.859 0.009***

Body Mass index <18.5 1.596 0.224 18.5-22 1.000 - >=22 0.681 0.273

Time since marriage 1.560 0.000*** Tim ef 0.991 0.000***

Intercept (Coefficient) -5.321 0.000***

Total cases (months) 1763 Births (n) 94

Note: Reference category has odds ratio of 1.000 tp=<0.1, **p=<0.05,***p=<0.005

61 Chapter 4: Fertility

4.7 Analyses of inter-birth intervals

‘Women give birth aii day and night women are taadhi, [fertiie] and birth-spacing is shorter. The time between the marriage and the first child is less than 1 year, previously it was around 3 years’ [Female informant, Daya Debeso]

For the analyses of birth spacing, two correlates of inter-birth interval length are included [for descriptions of dependent variables see section 4.3.2]: the length of the open birth interval and duration of post-partum amenorrhoea (date of return to menses).

4.7.1 Univariate analyses The general characteristics of the sample of Oromo women are shown in Table 4.6. Calculated using univariate life-table techniques, the overall median length of post­ partum amenorrhoea is 31.60 months and the median duration of breast-feeding is 37.50. A median open birth interval of nearly 19 months gives an indication of this sample’s high fecundability. In a comparable rural Indian population (Nath, Pers. Comm) found an open birth interval of over 26. However, in both studies these figures do not represent the characteristics of the entire population, since the samples are skewed towards less fecund women (pregnant women are excluded from the analysis).

4.7.2 Multivariate analyses Model 1 (Table 4.7) presents the results of a discrete time logistic regression model of the effects of socio-demographic variables and water tap installation (n= 893). In Model 2 (Table 4.8), a second model introduces additional covariates to control for women’s current nutritional status, body mass index and skinfold thickness, for a smaller sample (n= 324).

To test the assumption of proportional hazards, separate models were run (not shown here) to explore any interaction effects between length of exposure and the independent variables. There was no significant level of interaction, indicating that the assumption of proportional hazards is valid. An odds ratio greater than unity for time (months since last birth) and less than unity for time squared in all models indicates that both the risks of birth and return to menses have a curvilinear relationship with time.

62 Chapter 4: Fertility

Table 4.6 Description of the major variables for the analysis of length of post-partum

amenorrhoea and open birth intervals

Variable Definition N* Median or Mean (±) or % values Dependent variables Post partum Month of return to menses since 893 31.60 menses previous birth

Breast-feeding Full weaning from breast-feeding in 893 37.50 completed months

Length of open birth interval in 893 18.85 Open Interval completed months since previous birth

Independent variables Current age Current age of mother in completed 893 28.92 (±7.79) years [<20/20-29/30-39/40+]

Parity Number of live births 893 4.84 (±3.06) [1/2-3/4-6/7+]

BMI Body Mass Index: 324 20.30 (±1.93) (weight in kg*100)/height in cm^)

Skinfolds Sum of triceps and subscapular 324 19.01 (±5.95) skinfold thicknesses in mm

Preceding interval Length of last closed interval in 684 30.55 (±9.29) completed months [<24/24-46/47+]

Cattle Number of cattle owned by head of 893 3.32 (±4.04) household [None/1-5/ 6+]

A ltitude Village ecology - grouped by altitude 893 55.7% highland [Low/High] 44.3% lowland

Access to taps Village-level presence of water tap 893 37.7% live in tap stands villages [entered as time-varying covariate]

Survival status Survival status of previous birth 893 7.3% died before [time-varying covariate] survey date

Polygyny Marital status of head of household 893 23.4% polygynous [1 wife/Many wives]

Religion Religious status/ ethnicity of head of 893 73.6% Muslim household 26.4% Christian [Muslim/ Orthodox Christian]

Radio Household ownership of a radio 893 16.7% own a radio [Y e s/N o ]

Water carrier Method of water transportation 893 10% exclusively on [Always on back/ With assistance] back Notes; * Number of cases with non-missing values

63 Chapter 4: Fertility

In all models the observed relationships suggest that younger women (<20) have a shorter period of post-partum amenorrhoea and higher risk of childbirth than those from older age groups. There is no separate effect of parity on either return to menses or length of the open birth interval. Since early age at marriage is common among Oromo women (mean age= 15.85 ±2.00) these two variables are likely to be highly correlated.

The death of a child is associated with an early return to menses and a shorter birth interval length, and there is a significant level of interaction between the survival status of the child and length of exposure (time since birth). The negative coefficient in this interaction indicates that for women whose children die during infancy, the earlier the child dies during the interval the greater the risk of returning to menses and experiencing a subsequent birth. A positive coefficient for the interaction survival*time squared indicates that this effect has a curvilinear relationship with time, possibly an effect of the prolonged lactational amenorrhoea occurring while the child survives.

The effect of village ecology on open birth interval indicates that ecological conditions associated with altitude influence fertility differentials. Women living in the agriculturally fertile highland villages have a tendency to be exposed to a greater risk of short birth intervals than women living under the relatively harsh conditions of the lowlands, however these effects do not reach statistical significance (p=<0.05).

There are no clear relationships between fertility and measures of economic status. Radio ovmership is associated with a short open birth interval, but there is no effect on duration of post-partum amenorrhoea. There is no significant association between cattle ownership/herd size and fertility. Religion/ethnicity has no independent effect on rate of childbearing. However, the marital status of each woman’s husband may be a good predictor of birth spacing. Women in polygynous unions experience relatively longer birth intervals, possibly because of reduced coital frequencies. However, there is also a non-significant trend for these women to sustain longer periods of post-partum amenorrhoea, which may indicate some variation in breast-feeding practices or nutrition.

64 Chapter 4: Fertility

Model 1 (Table 4.7) clearly indicates that there is a positive effect of access to water taps on monthly risk of return to menses and childbirth. Improved access to water at any point since last birth increases a woman’s risk of returning to menses and experiencing a shorter birth interval. Furthermore, women who exclusively carry water on their backs are at a lower risk of returning to menses and have a tendency to experience longer birth intervals than women with no water-carrying load. A Kaplan-Meier survival curve in Figure 4.6 illustrates the increased monthly probability of returning to menses for women in villages with taps.

This effect is unlikely to be due to changes in childcare opportunities, since the duration of breast-feeding is not significantly longer in these villages (Table 4.9). Furthermore, there is no evidence of any improvement in nutritional status associated with the tap stands, since neither BMI nor skinfold measures are higher in these villages (Table 4.9).

Fig. 4.6 Effect of improved access to water points on the timing of return to menses following a birth 1.0

,8

.6

.4 Water source c o Taps o CL 2 .2 CL No taps 0.0 0 10 20 30 40 50 60

Months since birth

In Model 2 (Table 4.8) both measures of maternal nutritional status (when entered in the model independently or together) have no significant independent effect on return to menses, while all other effects remain constant. The higher categories of adiposity (sum

65 Chapter 4: Fertility of skinfolds >20) are associated with relatively short open birth interval lengths. Indicating that the women who gave birth most recently have higher skinfold measures. In a separate model (not presented here) the length of the preceding interval had no significant independent effect on the current birth interval or return to menses.

66 Chapter 4: Fertility

Table 4.7 Multivariate hazard regression model for length of post-partum amenorrhoea and open birth interval (Model 1) [| p=<0.1, ** p=<0.05, *** p=<0.005]

Variable Time to menses Open birth interval length Odds ratio P Odds ratio P Altitude: Lowland 1.000 - 1.000 - Highland 1.126 0.390 1.157 0 .0 6 1 t

W ater taps: No taps 1.000 - 1.000 - W ith taps 1.373 0.017** 1.171 0.044**

Parity:

1 1.000 - 1.000 - 2-3 0.161 0.539 0.977 0.977 4-6 0.167 0.566 0.401 0.401 7+ 0.513 0.182 0.958 0.815

Maternal age at birth: <20 1.783 0.024** 1.198 0.150 20-29 1.000 - 1.000 - 30-39 0.572 0.000*** 0.673 0.001*** 40+ 0.207 0.000*** 0.445 0.000***

Survival status of previous: Alive 1.000 - 1.000 - Dead 37.37 0.000*** 2.955 0.000***

Polygyny: Yes 0.764 0.0791 0.707 0.000*** No 1.000 - 1.000 -

Religion:

Muslim 1.000 - 1.000 - Christian 0.932 0.674 0.890 0.213

Radio owner: Yes 1.000 - 1.000 - No 0.994 0.975 0.794 0.021**

Cattle owner None 1.000 - 1.000 - 1-5 0.745 0 .0 7 2 f 1.057 0.564 6+ 0.760 0.150 0.960 0.747

Water carrier: Yes 0.602 0.021** 0.795 0 .0 6 2 t

No 1.000 - 1.000 -

Time (Months) 1.224 0.000*** 1.055 0.000*** Time^ 0.998 0.000*** 1.000 0 .0 5 4 t Survival*Time 0.817 0.000*** 0.873 0.000*** Survival*Time^ 1.002 0.001*** 1.002 0.000*** Intercept -6.498 0.000*** -2.793 0.000*** Person months 17071 16712 Events (n) 308

67 Chapter 4: Fertility

Table 4.8 Multivariate hazard regression model for length of post-partum amenorrhoea and open birth interval with anthropometric variables (other variables as Table 4.7, not shown) (Model 2)

Time to menses Open birth interval length

BMI (kg/m^': <18.5 0.707 0.266 1.277 0.144 18.5-22 1.000 - 1.000 - 22+ 0.615 0.235 1.448 0.0791

Sum of skinfolds

(mm): 1.000 -- - <20 0.999 0.959 1.420 0.014** 20+

Intercept -5.881 0.000*** -2.967 0.000*** Person months 6003 5766 Events (n) 114

Notes; tp=<0.1, **p=<0.05, ***p=<0.005

Table 4.9 Median duration of correlates of birth-spacing and mean measures of anthropometric status by village-level access to water tap stands.

Village access to water

Without taps W ith taps Median Median n - 557 n= 337

Return to menses 33.16 8.19 28.67** 3.44 (m ths)

Duration of breast­ 36.96 4.08 37 89 0.00 feeding (mths)

Open birth interval (m ths) 15.32 0.00 14.65 0.00 Mean Mean n= 190 n= 134

BMI (kg/m^) 20.48 ±2.02 20.06 ±1.76

Sum of skinfolds 19.03 ±6.37 18.98 ±5.34 (m m )

Notes: ** Wilcoxon test p= <0.05 ^ % still amenorrheic/breastfeeding after 60 mths

68 Chapter 4: Fertility

4.8 Discussion Overall the local response to the improved water-supply scheme has been very positive. Women state that they are now able to spend the time freed from water collection on less arduous activities within the home, and the children who once accompanied them are now free to attend school. However, the analyses presented here indicate that the installation of taps is also associated with a shift in the timing of reproductive events, towards increasing fertility. Changes in key reproductive events are observed in response to the introduction of water points, which are absent in adjacent villages without access to new water sources.

Firstly, mean age at marriage (and age at first birth) has increased. This finding is the reverse of the effect observed among the Maya, where adolescent workloads were reduced and median age at first birth dropped (Kramer and McMillan 1999). The shift towards later age at marriage among the Oromo women may be to be related to secular trends occurring in the region, most notably through the growth of regional market towns which have provided access to improved information and schools. A further explanation may relate to a trend towards later age at menstruation, which would influence the age at which a young girl is considered marriageable (Graham, Larsen et al. 1999; Ramakrishnan, Barnhart et al. 1999). This might occur under conditions of declining food availability and poor childhood nutrition.

Additionally, there is an independent finding that the new taps stands are associated with shorter birth intervals. Multivariate hazards analyses demonstrate that, after controlling for secular trends in socio-demographic variables, women who have access to the improved water supply are experiencing earlier first births following marriage and shorter subsequent birth interval lengths, associated with a reduced period of amenorrhoea following childbirth.

4.8.1 Change in nutritional status? Other studies have highlighted the effects of nutritional status on the duration of the birth interval. Undemutrition is associated with: delaying reproductive maturation (Foster, Menken et al. 1986), prolonging lactatational anovulation (Huffman 1987); affecting ovarian function (elevating risk of anovulatory cycles, oligomenorrhoea or

69 Chapter 4: Fertility luteal insufficiency) (Lager and Ellison 1990; Cumming 1993) and increasing of the risk of early intrauterine mortality (Ford, Huffman et al. 1989); which serve to lengthen birth interval lengths. However, the results of the multivariate hazards models presented here suggest that maternal nutritional status is not the strongest predictor of monthly risk of first birth for the recently married sample of women. In fact there is a tendency for poorly-nourished women to have the shortest first birth intervals. Additionally, nutritional status also has no relationship with the resumption of menstruation following subsequent births. These results add to a growing body of research suggesting that the effects of subtle forms of undemutrition may not influence fertility (Delgado, Martorell et al. 1982; John, Menken et al. 1987; Strassman and Warner 1998).

4.8.2 Altered breastfeeding practices? Lactational infecundability has been cited as being one of the most important regulators of fertility in non-contracepting populations (Bongaarts and Potter 1983) through the modulation of the suckling-stimulated secretion of prolactin and the alteration of hormone release pattern in the pituitary gland (Vitzthum 1997). Among the Oromo, supplementary foods, including porridge (barley, milk and butter), are introduced around 6 months of age and full weaning is unlikely to take place until the next pregnancy. In the absence of detailed behavioural information on the intensity and frequency of breast-feeding, breast-feeding duration (or time to full-weaning) is used as a crude indicator of the relative breast-feeding practices after birth. Table 4.8 indicates that there has been no change in breast-feeding lengths associated with the installation of taps, since there is no variation across villages. However, it is difficult to interpret the exactly what role is played by lactation in determining fertility in this population, since many women resume menses while still nursing and may continue to breastfeed until the next birth.

4.8.3 Reduced disease loads? An alternative explanation for shorter patterns of birth-spacing could relate to a reduction in women’s parasite loads and a general improvement in health associated with the improved water supply. However, the results of a two-week recall women’s health survey indicate no variation in illness prevalence for women with access to improved water compared with those without [described in Chapter 5].

70 Chapter 4: Fertility

4.8.4 Reduced energetic expenditure? A reduction in women’s workloads may explain the shorter pattern of birth-spacing observed in the analysis [outlined in Chapter 1: Section 1.4.1]. There is no evidence to suggest that access to taps is associated with an improvement in women’s total energy balance [energy input - energy expenditure] (based on the assumption that the traditional water source villages are a proxy for conditions prior to the installation of taps (Table 4.9)). However, clinical studies indicate that a reduction in energy expenditure independent of weight change may influence reproductive function (Shangold, Freeman et al. 1979; Beitens, McArthur et al. 1991; Jasienska and Ellison 1998). Although there are no hormonal data available for this Oromo population, if reproductive function responds in the same way then the reduction in women’s water- carrying workload may explain patterns of shorter birth spacing observed in this population. Reduced workloads may have also contributed to a reduction in early foetal loss, as pregnancies are successfully brought to full-term. However, information on conception frequencies and rates of pregnancy loss cannot be observed in the demographic data set [Chapter 5: Section 5.2.3]. Additionally, the independent finding, that there is a tendency for women water-carriers to be exposed to lower risks of birth per month than women who are assisted with water-carrying, suggests that the energetic costs of water-carrying are great enough to influence fertility.

The modulation of ovarian function and foetal loss in response to changes in energy levels among women has adaptive significance. Such a flexible response system could have served to prevent the wasteful allocation of resources to reproduction under stressful environmental conditions, ensuring the survival of women and the maintenance of future reproductive potential. However, this mechanism may operate even for individuals who are not in a state of negative energy balance. Two adaptive explanations have been proposed. First, simply that a change in energy expenditure serves as an indicator of an imminent change in energetic status (balance). Additionally, that high energy-expenditure may conflict with energy-sparing mechanisms normally mobilised during pregnancy, e.g. a reduction in basal metabolic rate (Jasienska 2001).

71 Chapter 4: Fertility

4.8.5 Increased coital frequency? A final explanation for the shorter birth spacing observed in the analyses, may relate to behavioural changes associated with an increase in coital frequencies. Since the introduction of new tap stands is associated with more free time for women spent around the homestead, it is possible that the reduction in women’s workloads may either improve marital relations or facilitate increased coital frequencies. In the absence of data on the frequency of sexual activity, it is only possible to speculate about the likelihood of increased sexual activity. However, since male workloads have been unaffected by the new village-level tap stands, it seems unlikely that there would be any greater opportunity for sexual relations.

4.9 Concluding remarks

Overall the results indicate that the installation of village-level tap stands has shortened birth intervals independent of other socio-demographic changes occurring in the region. This effect is likely to be mediated by variation in patterns of energetic workloads, since women’s nutrition, health and breast-feeding practices have not changed. Since the timing of first birth and the pace of later birth scheduling bears a negative relationship with family size in non-contracepting populations (Bumpass, Rindfuss et al. 1978; Trussell and Menken 1978), the introduction of a new water supply in Hitosa and Dodota districts may result in unforeseen population growth. The consequences of shorter birth-spacing patterns for maternal [Chapter 5] and child [Chapter 6] health and well-being are explored in following chapters.

72 5

Women’s health and body condition - evidence for maternal depletion

‘ P ro b le m s [o f ill health/ begin in the mother’s stomach. Mothers are weaker now.’ [Male Informant, DS]

5.1 The status o f women Numerous studies have documented evidence of a worsening of women’s well-being in transitional rural communities in relation to autonomy, workload, literacy, nutrition and disease prevalence (Raikest 1989; Kunstadter 2001). New eco-demographic changes are associated with an increase in male out-migration and the marginalisation of rural communities. Mothers are among the most vulnerable members of these communities, since they suffer the extra energetic costs of childbearing.

Maternal energetic deficiency and malnutrition is commonly reported in Ethiopian populations (Wenlock and Wenlock 1981; Berhane, Gossaye et al. 2001) attributed to the effects of high fertility levels, heavy workload and gender biased intra-household food allocation (women generally eating last and least). Moreover, traditional harmful practices (e.g. female genital mutilation, early age at marriage), high rates of infection and low access to health and education place the health of women at risk. Maternal mortality rates are among the highest in the world: 560-850 women dying in childbirth per 100,000 live births (CSA, 2001). Higher mortality in middle-aged women compared to men has also been observed in this region (Berhane 2000).

In this chapter I shall outline the health and nutrition of women in the villages of Hitosa and Dodota and then examine the consequences of the access to improved water supplies and higher fertility levels for maternal condition. It is predicted that in the absence of adequate health care facilities and low food availability, closer birth spacing Chapters: Women’s condition

associated with the labour-saving technology may impose severe constraints on matemal-child health through maternal depletion syndrome. ‘Maternal depletion syndrome’ is a the term for the deleterious effects of repeated pregnancies, closely spaced births and extended periods of breast-feeding which result in a general decline in maternal health and condition. Poor maternal body condition may also contribute to high prenatal and neonatal mortality [described in Chapter 6].

5.2 Health characteristics

The health characteristics of all ever-married women contacted during the household demographic survey are outlined in Table 5.1. A total of 1566 women in seven villages completed the health survey [described in Chapter 2: Section 2.6]. Overall 15.2% of the women reported that they were ill on the day of interview, and over one quarter of them experienced illness at one point during the two weeks preceding the interview. For those women reporting at least one bout of ill health during the two week recall period (n= 449), 62.6% said their illness confined them to bed during the day and 34.1% were ill for a time period of over one week. The average duration of each spell of illness was 5 days.

Table 6.1 Health data from all ever-married women by village

Daya Daya Terro Hurturbe Reissa Debula Bekare A ll Debeso Gabrel Moyee Michiko Saapo Washo Sample 186 279 196 171 211 272 251 1566 size

III on day 23.7 15.4 17.9 6.4 15.2 12.5 15.5 15.2 %

III over 2 w eeks % 29 27.1 25.5 16.4 32.7 33.1 33.1 28.7

Lay down 50 54.7 66 75 66.7 62.2 68.7 62.6

Days III Mean 3.94 3.45 6.04 5.46 4.78 7.13 6.47 5.43 ±SD ±7.09 ± 5.04 ±7.31 ±6.69 ± 5.58 ± 15.63 ±8.09 ±9.26

None im m un. 23.7 24.7 25.5 16.4 26.1 31.6 27.5 25.6

i. 1 Notes: Women whose illness forced them to lie down during the day (severity of illness) ^Women with none of their children immunised

74 Chapters: Women’s condition

The most notable difference in health reporting between the villages related to one of the most remote highland villages without access to water development, Hurturbe [Chapter 3: Table 3.1]. Here fewer women reported illness on the survey date or during the 2-week recall period, and fewer women had left their children un-immunised than in any other village. This may, in part, relate to the villagers’ greater wealth, demonstrated by larger household herd sizes (Table 5.2). Milk products may also provide a valuable source of nutrients for these households.

Table 5.2 Indices of village wealth

Village Altitude Herd size (mean ± SO) Radio ownership (%)

Daya Debeso High 2.22 ± 1.95 21.5 Daya Gabrel High 2.24 ± 2.61 24.8 Terro Moyee High 5.10 ±4.49 23.3 Hurturbe High 5.15 ± 6.22 23.1 Reissa Michiko Low 3.33 ± 3.30 6.7 Debuia Saapo Low 2.64 ± 2.68 7.8 Bekare Washo Low 3.39 ± 2.57 6.9

5.2.1 Health seeking behaviour In this region health care facilities are poorly distributed; access to clinics is confined to local market towns, which are over 3 hours on foot from any village surveyed. Many drugs (e.g. anthelminthics, antibiotics) and hospital facilities are only available in the zonal capital in Asella over 30km away and can be costly. Accordingly, the majority of the women surveyed reported not seeking any treatment for their illness (60.1%), a small percentage treated themselves with traditional remedies (6%), while only 34% had attended a clinic.

Improved communications and the construction of new roads in the early 1990s has enabled outreach clinics to attend villages annually to inoculate new-boms against Polio, DPT (diphtheria, whooping cough and tetanus), BCG (tuberculosis) and measles, and to administer tetanus vaccinations to pregnant women. However, despite access to outreach immunisation programmes, one quarter of women surveyed stated they had not inoculated any of their children against any of these diseases.

75 Chapters: Women’s condition

5.2.2 Reported illness When questioned regarding the symptoms of their illness, the most commonly reported ailments related to intestinal complaints (21.9%), including diarrhoea, vomiting and stomach cramps), which were associated with eating unripe grain and with parasites. Severe muscle/back pain linked with undertaking heavy household chores (21.7%), e.g. collecting water and firewood, was also prevalent. Respiratory tract infections were described by 12.2% of women (e.g. fevers and coughing, tuberculosis) and were attributed to working outdoors. Eye infections were reported by 10.3% of women. A number of gynaecological problems, including heavy menstrual bleeding, and obstetric complaints were reported by 5.4%. Treatment for these ‘female conditions’ was generally only sought from traditional healers. Few women reported suffering from anaemia (3.3%), however this complaint may be widespread. A local health care professional estimated that 75% of the female patients attending a nearby clinic suffered from anaemia, a problem exacerbated by poor diet, close-birth spacing and malaria in low-lying areas.

5.2.3 Miscarriages and still births During demographic data collection every effort was made to record the details of all failed pregnancies (miscarriages and still births); however only 2% of women reported experiencing at least one miscarriage and 1.5% reported a still birth. These figures are not likely to represent an accurate level of naturally aborted pregnancies, since most miscarriages occur in the first trimester when women are not aware that they are pregnant. Without hormonal data on conception dates it is difficult to calculate levels of intra-uterine foetal loss. Furthermore, while efforts were made during data collection to ensure data quality [described in Chapter 2], recall biases and a reluctance to mention failed pregnancies may have prevented women from clearly reporting the details of non- live births. In the health survey one common health concern of women related to ‘an overflow of menstruation’, which may indicate the occurrence of a spontaneous abortion. There are no data to accurately calculate maternal mortality levels for this population; however, 5% of children aged 0-14 enumerated in the household survey had experienced the death of their biological mother.

76 Chapters: Women’s condition

5.3 Levels of energetic status

Body size and composition data may be used to describe nutritional status; however, it is better to consider them as a broader measure of maternal condition or energetic status rather than purely of dietary intake. Health, workloads and reproductive scheduling are all likely to influence energy levels, the utilisation of body stores and growth. The characteristics of the sample measured during the post-harvest dry season are comparable with levels identified in other regional studies (Table 5.3) in which evidence of both chronic and transient forms of energy deficiency have been documented (Ferro-Luzzi 1990; Branca 1993). In the villages of Hitosa and Dodota woredas 21.8% of the married women surveyed had a body mass index below the critical cut-off for chronic energy malnutrition (<18.5).

Table 5.3 Anthropometric data from women in Southern Ethiopia

Hitosa/Dodota^ Oromiya^ Sidama Age 31.65 ±9.97 15-49 38 ± 9 ' Height (cm) 156.17 ±5.68 156.8 154 ± 6 ' Weight (kg) 49.16 ± 5.95 44.5 ± 6.2' BMI 20.14 ± 1.98 19.8 19.0 ± 1.6' % < 1 8 .5 21.8 28.7 16' MUAC (cm) 24.51 ±2.15 23.8 ± 1.1' Triceps (mm) 10.18 ±3.58 8.9 ± 2.3' Subscapular (mm) 8.56 ± 2.84 10.4 ± 4 .2 ' Note: ever married women <50 (n=455)’ CSA (DHS) 2001(n=5121); ^Taffesse 2001 (n=225); ^Ferro-Luzzi et al. 1990 (n=33)

Regional dietary intakes are characterised by very high carbohydrate, and low fat and animal protein, which are inconsistent with the recommended levels required for optimal health and reproductive performance (Selinus, Gobezie et al. 1971; Taffesse and Girma 2001). A report evaluating the nutritional concerns of Arsi in 1999 identified acute malnutrition among both women and children (Belay 1999) and recommended food aid to be distributed in a number of districts.

Evidence documenting high fertility levels and the water collection workloads experienced by women in this region are outlined in Chapter 4 and Chapter 3 (Table 3.5) respectively. However, women also undertake other labour-intensive household chores, which include firewood collection, grinding maize and transporting produce to/from the market. Furthermore, female household-heads (9.2% of households) face additional responsibilities for all economic activities in the absence of an adult male.

77 Chapters: Women’s condition

Women are likely to sustain considerable energetic stresses associated with combining both reproductive and productive roles.

Table 5.4 documents the anthropometric characteristics of women by village. Performing univariate ANOVA statistical analyses and Posthoc LSD tests across the villages reveals that only women from Daya Debeso have statistically higher anthropometric measures (height and body fat) compared to women from the lowland villages (Reissa Michiko, Debula Saapo). This may relate the poorer conditions associated with low-lying villages, including less fertile lands, which support fewer cash crops.

Table 5.4 Anthropometric characteristics of ever-married women (<50) by village

Daya Daya Terro Hurturbe Reissa Debula A ll Debeso Gabrel Moyee M ichiko Saapo Sample 92 86 45 85 85 62 455 Age (yr) Mean 32.87 31.80 33.33 30.12 30.75 31.95 31.68 ±SD ± 10.52 ± 9.45 ± 9.51 ± 9.83 ± 9.89 ± 10.17 ± 9 .9 4 Parity Mean 5.53 5.74 6.67 5.16 4.32 5.02 5.32 ±SD ±3.54 ±3.24 ±3.21 ±3.44 ±2.92 ±3.21 ± 3 .3 2 Height(cm)** Mean 158.38 155.82 155.89 155.36 155.71 155.31 156.17 ±SD ± 5.47 ±5.53 ± 5.01 ± 5.95 ± 5.96 ±5.29 ± 5.68 Weight (kg) Mean 50.44 47.44 49.15 49.45 49.53 48.76 49.16 ±SD ± 5.66 ± 5.66 ±5.80 ±5.84 ± 6.56 ±5.81 ± 5 .9 5 BMI (kg/m^) Mean 20.09 19.53 20.22 20.48 20.37 20.18 20.14 ±SD ± 1.91 ±2.04 ±2.15 ± 1.99 ± 1.89 ± 1.83 ± 1 .9 8 %<18.5 22.8 31.8 18.4 22.4 18.8 21.0 21.8 Tricep** Mean 11.33 9.51 10.07 10.31 9.89 9.71 10.18 ±SD ±3.73 ±3.25 ±4.20 ±4.00 ±3.04 ± 3.06 ± 3.58 Subscap** Mean 9.59 7.91 8.47 8.77 8.41 7.90 8.56 ±SD ±3.17 ±2.67 ± 3.05 ±3.26 ±2.10 ±2.17 ± 2 .8 4 MUAC (cm) Mean 24.94 23.83 24.46 24.74 24.65 24.28 24.51 ±SD ±2.44 ± 1.99 ±2.11 ±2.89 ± 1.96 ± 1.96 ± 2 .1 5 Notes: ** Significant (<0.05) Post Hoc LSD tests: Height - Daya Debeso * all villages; Triceps - DD * DG, RM, DS; Subscap. - DD * all villages

5.4 Does improved water supply improve maternai condition? At the survey date there is no evidence indicating that villages with access to the new water development projects have improved maternal condition. Univariate T-test

78 Chapters: Women’s condition

analyses reveal that the mean skinfolds measure of body fat does not differ significantly between women in villages with access to improved water supplies compared with those villages without; furthermore, body mass indices may be higher in villagers without access to the new water supply (Table 5.5).

Table 5.5 Health and nutritional status of women by village water source

Village access to water Without taps With taps Sample (n) 215 240

BMI Mean(kg/m^)^ 20.38 ±1.98 19.92 ±1.95

Sum of 2 skinfolds Mean (mm) ^______18.66 ______±6.04 ______18.81^______±5.90 Sample (n) 829 735

% reported Illness 27.7% 29.8% (2 week recall)

% treated at a clinic ^______37%______30.6% Note: h=2.50, p=0.13** V-0.28, not significant (>0.05) Vo women reporting illness (n=447)

Similar univariate analyses of the health survey data indicate there is no variation in illness prevalence or health-seeking behaviour between villages categorised by water supply access (Table 5.5). Since villages included in the study were selected for comparability in all aspects except water supply (e.g. size, altitude, proximity to healthcare) [Chapter 2: Section 2.1.1], this result indicates that the new water supply project has not independently improved women’s health. However, all the villages are likely to have benefited from improvements in regional health care services since the early 1990s.

5.5 Does closer birth-spacing effect maternai condition ? While the presence of the water technology is not associated with any improvements in anthropometric status among women, it is linked with shorter birth spacing [Chapter 4]. In this section I examine whether an increase in reproductive energetic stresses associated with these higher fertility levels may be detrimental for maternal health and well-being.

79 Chapters: Women’s condition

5.5.1 Energetic costs of reproduction Reproduction exerts energetic stress on mothers. Cumulative stresses imposed by successive pregnancies and lactation may result in ‘maternal depletion syndrome’ leading to maternal weight loss, anaemia, obstetric complaints and low birth weight offspring (Winikoff and Castle 1988; Rawlings, Rawlings et al. 1995; Khan, Chien et al. 1998). Maternal depletion involves the incomplete restoration of physiological nutritional reserves that are depleted in the course of pregnancy and lactation (Winkvist, Rasmussen et al. 1992); women subsidise the costs of successive pregnancies through a gradual deterioration in their own nutritional status. The energy depleting effects of reproduction are frequently evaluated by testing for an association between parity or birth interval length as a proxy measure of maternal and perinatal condition, e.g. weight, skinfold, morbidity and mortality.

While some studies show an inverse relationship between parity and maternal weight in chronically malnourished populations (Chowdhury 1987; Tracer 1991; Little, Leslie et al. 1992; Gamer, Smith et al. 1994), others show an opposing association between high parity and improved energetic status, or no association at all (Prentice, Whitehead et al. 1981; Miller and Huss-Ashmore 1989). However, a methodological problem associated with using parity as an indicator of reproductive stress is that high parity women represent healthier individuals in the population. A selection bias towards well- conditioned mothers at higher parities may have concealed any maternal depletion present in these studies, i.e. phenotypic correlations.

Research undertaken using birth intervals as the measure of stress also produce conflicting results. Short intervals between pregnancies are associated with several unfavourable pregnancy outcomes, e.g. increased prevalence of low birth weight babies (Ferraz, Gray et al. 1988; Miller 1989; Huttly, Victora et al. 1992), increased intra­ uterine (Hebert, Bouyer et al. 1986) and perinatal mortality (Fedrick and Adelstein 1973). However, not all researchers have been able to verify the existence of a negative change in maternal nutritional status (Winkvist, Jalil et al. 1994). Defined as a negative change in maternal nutritional status within a single reproductive cycle, maternal depletion is most likely to occur only among women with a marginally inadequate food intake and an inability to make behavioural and thus energy expenditure adjustments to low intake.

80 Chapter 5: Women’s condition

Under extreme (i.e. unpredictable) conditions of malnutrition, for which adaptations may not have developed (e.g. the Dutch Famine during WW2) when energy intake declines during the third trimester, both maternal postpartum weight and infant birth weight decrease with parity (Stein, Susser et al. 1975). However, under conditions of moderate malnutrition women are able to replete themselves during successive rounds of reproduction (Adair 1992; Miller, Rodriguez et al. 1994), but with concurrent negative effects on offspring weight (Winkvist, Habicht et al. 1998). In other words, among very poor conditioned individuals there is preferential partitioning of nutrients, which may protect the mother at the expense of her foetus. However, there is a reverse effect among better conditioned (but still marginally nourished) women, who lose weight, while their infants birth weights are protected (Merchant, Martorell et al. 1990; Winkvist, Habicht et al. 1998). However, higher pregnancy weight gain among underweight Austrian women did not compensate for the negative impact of poor pre­ pregnancy weight status (Kirchengast and Hartmann 1998). Overall these results indicate that even in highly developed populations poor maternal condition maybe a risk factor for maternal weight loss, growth retardation and/or low weight offspring.

5.6 Analyses of the main determinants of maternai condition To examine the effect of reproduction on women’s condition, and to identify the major correlates of maternal energetic status and good health in the Oromo sample, two models of multivariate regression analyses are performed. Table 5.6 describes the major correlate of two measures of women’s body condition using multiple linear regression analyses and Table 5.7 describes the major correlates of two measures of women’s good health using multiple logistic regression analyses.

In Table 5.6 two measures of body condition: body mass indices and the sum of two skinfolds (triceps and subscapular) are included as the dependent variables. Although the nutritional status of each woman is likely to fluctuate between seasons, these current status measures are used as an overall indicator of variation between women during the dry season months. Logarithmic transformations are performed to correct for skewness in the skinfolds variable (a simple measure of percentage body fat). In this case all ever- married women of reproductive age (<50 years) included in the anthropometric survey

81 Chapters: Women’s condition who had experienced at least two births within the 6 year observation period (n= 261) are included in the analysis. The mean age of the women included in the sample is 29.5 ± 6.99.

Two measures of well-being, good health on the day of survey and good health during the two week recall period, are the dependant variables described in the second model presented in Table 5.7. In this case all reproductive-aged married women (<50 years) who had experienced at least one birth within the 6 year observation period (n= 1100) are included in the analysis.

Categories of covariates initially entered into both models include both socio­ economic/ecological variables [altitude, herd size, radio ownership, religion, level o f education, locality, method o f water collection] and demographic variables [age, parity, survival status of the most recent birth, time elapsed since birth as a linear and quadratic function, length o f preceding full birth interval]. Variables consistently non-significant with the lowest parameter estimates are excluded from the final models presented here.

5.6.1 Correlates of body condition The results of multivariate regression analyses performed to isolate the main correlates of maternal energetic demonstrate that neither the socio-ecological covariates nor access to tap stands has any independent effect on a woman’s condition (BMI and adiposity) (Table 5.6). However, demographic characteristics explain the greatest part of the observed variation in anthropometric measures.

The time elapsed since last birth is one of the strongest correlates of body condition. Mothers who gave birth recently have the highest BMI (Fig 5.1) and skinfold values (Fig 5.2). For the skinfolds measure this is expressed as a curvilinear relationship across the birth interval, indicating that fat reserves are gradually restored before the next reproductive cycle. Lengthening this period of recuperation after weaning has been shown to improve women’s nutritional status before the next pregnancy (Adair and Popkin 1992).

Length of the preceding interval is an indicator of a woman’s recent birth-spacing pattern and is inversely related to current energetic status, indicating that women

82 Chapters: Women’s condition experiencing longer birth intervals may be in worse condition during the early stages of the next round of reproduction. This may relate to the high energetic costs of prolonging lactation for women continuing to breast-feed up to the next pregnancy (Rashid and Ulijaszek 1999; Adair and Popkin 1992); however, this outcome may also reflect the presence of a group of highly stressed mothers who experience consecutive long birth intervals due to unmeasured energetic stresses, e.g. low food intake.

There is a relationship between survival status of the previous child and maternal condition. Women whose previous child had died before the survey date have lower skinfolds measure than women whose child survived the interval; however this does not quite reach statistical significance (p=<0.05). Similarly, the effect may signify that poorly nourished women are at greater risk of experiencing an infant death, being unable to sustain the costs of lactation and childcare.

Fig 5.1 Scatter plot of body mass index by time since last birth

28

26

24

22 ■ ■ ■ ■ 20

18

16

■D O 14 10 20 30 40 50 60

Months since birth

83 Chapters: Women’s condition

Fig 5.2 Scatter plot of sum of two skinfolds by time since last birth

1.8

1.6 (/) ■ a

1.4

Ü3 ■ 'o ■■ ■ E 3 C/3 1.2 O0 3

1.0

.8 0 20 40 60 80

Months since birth

Body mass index is sensitive to both age and parity. There is a near significant positive relationship between BMI and age, which is likely to be due to an overlap between marriage and the period of adolescent growth (9% of the women in the sample are <20 years). There is also an independent negative effect of parity on body mass index. To interpret the relationship between nutritional status and parity, whilst controlling for the effects of age, the means of BMI stratified by parity and decade of life are presented in Figure 5.3. The strongest trend is across parity. Within each age-class individuals of higher parity have a lower BMI than their lower parity counterparts with the exception of <20 age-class, a group that will include those women married while still undergoing a phase of adolescent growth.

5.6.2 Correlates of health The results of multivariate regression analyses performed to isolate the main correlates of women’s health demonstrate that there is no significant health advantage afforded to women living in the villages with access to taps. There is a weak positive association between herd size and good health on the survey date (Table 5.7). Altitude is the best determinant of sustaining good health over the 2-week recall period; women from lowland villages are at a greater likelihood of experiencing a bout of ill health. Among

84 Chapters: Women’s condition

the demographic covariates, time lapsed since last birth has a strong negative relationship with sustaining good health across the 2-week period. In this model higher parity is associated with an increased likelihood of ill health on the survey day; however this effect does not quite reach statistical significance (p=<0.05), possibly due to small sample sizes. This outcome lends support to the prediction that large numbers of offspring may negatively affect maternal well-being.

Table 5.6 Multiple general linear regression model for correlates of body mass index and sum of skinfolds for reproductive-aged women (<50) with at least two births (n=261) Body mass index Sum of two skinfolds"’ Coefficient SE p Coefficient SE p Access to taps Taps 1.037 1.68 0.630 -0.103 0.105 0.330

No taps -- Altitude

Highland(n=192) -- Lowland (n=69) -1.951 2.48 0.432 7.68 0.159 0.631 Radio Yes (n=54) -2.38 1.85 0.20 4.102 0.119 0.731

No (n=207) -- Herd size -2.44 0.03 0.41 -4.33 0.002 0.820

Age 8.85 0.05 0.066t -1.07 0.003 0.972

Parity -0.22 1.11 0.046** -2.91 0.007 0.683

Survival status Alive (n=247) 1.66 1.01 0.10 0.119 0.07 0.07f

Dead (n=14) -- Preceding interval -3.48 0.02 0.02** -1.93 0.001 0.04**

Time since last birth (months) -6.62 0.03 0.03** -4.95 0.002 0.01** Time^ 8.55 0.00 0.201 8.62 0.00 0.04** Intercept 19.89 1.22 0.00*** 1.296 0.08 0.00*** R" 0.157 0.145 Corrected F 1.711 0.02** 1.533 0.047**

Fig 5.3 Mean body mass index (BMI) by parity and decade of life

20 . Maternal age

20.6 - -<20 20-29 20.4 30-39

CM E 20.2 O)

S 20 - - - CO

19.6

19.4 1 2-3 4-5 6+ Parity Chapters: Women’s condition

Table 5.7 Multiple logistic regression model for covariates of reported health for all reproductive aged women (<50) with a birth (n=1100)

W ell today Without illness during 2 weeks Coefficient SE P Coefficient SE P Access to taps Taps -0.128 0.19 .500 -0122 0.15 0.425

No taps - - A ltitude

Highland -- Lowland -0.044 0.19 0.815 -0.367 0.15 0.016** Radio Yes -1.94 0.25 0.439 -0.109 0.21 0.597 No -- Herd size 0.047 0.03 0.09f 0.015 0.02 0.461

Age -0.005 0.03 0.863 -0.017 0.02 0.420

Parity -0.105 0.06 0.09t -0.009 0.05 0.858

Time since last 0.008 0.00 0.899 -0.01 0.005 0.036** birth (months)

Intercept 2.491 0.52 0.00** 2.165 0.412 0.00*** Notes: ' log transformed, fp=<0.1, **p=<0.05, ***p=<0.005

5.7 Concluding remarks The general health and nutritional characteristics of the sample of Oromo women indicate that they are exposed to chronic energetic deficiencies, which are likely to be deleterious for mothers and may consequently affect child well-being. These deficiencies are caused by a combination of factors including increasing population pressures, repeated crop failures (environmental degradation), harmful cultural practices, e.g. early age at marriage and female genital mutilation, and inadequate developmental infrastructure (schools, healthcare services), which characterise many areas of rural Ethiopia.

Villages with access to an improved water supply do not have higher levels of women’s body condition or improved health than villages continuing to use traditional water sources. Assuming that these villages serve as a proxy for conditions before and after water taps were installed, this result indicates that maternal condition may not have been dramatically improved by access to the development scheme.

However, additional analyses reveal that a woman’s pattern of childbearing does strongly affect her body condition. There is evidence that maternal depletion in

86 Chapters: Women’s condition nutritional status, associated with a parity-specific decline in body fat, is prevalent. Depletion effects are likely to be apparent in this population because of a combination of the high fertility and heavy workloads, compounded by an underlying level of inadequate nutrition, which result in particularly severe energetic stress being placed on women. Overall these findings highlight the need for improved maternal health services in this region.

87 6

Determinants of child mortality

‘In the past, when a woman gave birth there was plenty of meat for the mother and [breast] milk for the child. Butter was placed near the child and It managed to sleep all day. Today there Is not much milk from women’s breasts, so water, not milk. Is given Immediately after birth and the child Is restless.’ [Female informant, Debula Saapo]

‘The main problem now for children Is malnutrition. Children don’t get a balanced diet since there Is no grazing land for cows, and even If there Is a cow It doesn’t produce enough milk.’[Male Informant, Daya Debeso]

‘Early child death Is caused by ‘tonclle’ ....the sym ptom s are diarrhoea, fever, depressed fontanelle. Inability to breastfeed.’ [Female informant, Daya Gabrel]

6 .1 Levels of mortality Levels of childhood mortality are responsive to changes in the local ecology, e.g. disease and nutrition levels, and therefore serve as a reliable indicator of the condition of the entire community. By analysing birth history data collected during the demographic survey of villages in Hitosa and Dodota districts (woredas), this chapter includes sections describing early childhood mortality at the survey date, fluctuations in mortality patterns over time, and exploring the major correlates of early child death. This includes an examination of the impact of recent development technologies, as well as other significant ecological events, which have occurred in these villages.

6.1.1 Period mortality rates Indirect techniques are the most widely used method for estimated levels of mortality from birth-history data. In this case the proportion of children ever bom who have died are used as indicators of child mortality for all women surveyed, producing total sample of 7795 live births. This method calculates a ratio, not a rate, since the Chapter 6: Child mortality denominator population is not the true population at risk of the event in the numerator, i.e. some deaths may be of births that occurred in the previous year.

Infant/child infant deaths (<1 years) or child deaths (<5vears) x 1000

mortality ratio = total live births in year

Mortality rates from a number of recent demographic surveys in Southern Ethiopia are described in Table 6.1. In the villages of Hitosa and Dodota woredas the infant mortality (iqo) is 143.6/1000 live births, indicating higher levels than those recorded for the wider region of Oromiya (116.2/1000) described in the National Demographic and Health Survey undertaken in 2000 (CSA 2001). Levels of neonatal mortality (<1 month) are very high for this population; over half of all infant deaths (54%) occur in the neonatal period. Later childhood mortality is comparable with the other regional studies (88/1000 live births).

Overall in Hitosa and Dodota, 8% of all live births die during the first month of life, 23% die during childhood (before the 5^^ year of life). Clearly young children are exposed to high mortality risks. Total under-five child mortality is 232.2/1000 live births, higher than both the regional and national averages.

Table 6.1: Data on early childhood morta ity in Southern Ethiopia Site Date NMRIMR CMR UnderS Reference (iqo) (4qi) (sdo) Hitosa/Dodota 1999/00 78.5 143.6 88.8 232.3 Arsi 1991 173 253 (Hailemariam 1991) Butajira, Sidamo 1993 136 293 (Shamebo 1993) Oromiya 1990-00 61.1 116.2 87.9 193.8 CSA (DHS) 2001 Ethiopia 1996-00 48.7 97 76.7 166.2 CSA (DHS) 2001 Notes: NMR - neonatal mortality (<1mth

6.1.2 Fluctuations in mortality overtime The entire data-set of live births obtained from all reproductive aged (<50) ever-married women is used to calculate age-specific mortality across a 30 year time period (n=7795). Age-specific mortality is calculated as the proportion of children dead classified by children’s two-year age groupings, which provides an estimate of the probabilities of dying between birth and a given childhood age.

Age specific mortality = deaths aged x in a 2 year group

children alive aged x at start of 2 year group

89 Chapter 6: Child mortality

Early childhood mortality has fluctuated considerably across the 20 years preceding the survey (1979-1999) (Fig. 6.1). Infant (<= 1 year) and later child deaths (age 1-5 years) have followed a similar pattern of high but gently declining mortality rates across this time period. The main trends within this period indicate an increase during the 1980s, which peak in the late 1980’s, followed by a slow but steady decline in mortality throughout the 1990s.

Fig 6.1 Age-specific mortality for all live births 1979-2000 (n=6382). [ — - ■ = infant (12CI0), — = later child ( 48 C113)]

w 250 I 200 § 150 % 100

79- 81- 83- 85- 87- 89- 91- 93- 95- 97- 99- 80 82 84 86 88 90 92 94 96 98 00

The trends in early childhood mortality overtime may relate to three important social and ecological events which occurred in the region. Firstly, the countrywide drought and famine and the political upheaval of the mid/late 1980s imposed conditions of severe energetic stress in rural households. Dramatic increases in early childhood mortality have been described across Ethiopia between 1984-1986 (Kidane 1989; Kidane 1990; Kloos and Lindtjorn 1994) and globally during other periods of famine [in South Asia: (Bongaarts and Cain 1982; Dyson 1991), North Korea: (Goodkind and West 2001) and Europe: (Stein, Susser et al. 1975)].

Secondly, the introduction of vaccination programmes in Arsi during the 1990s and the construction of government clinics are likely to have reduced the prevalence of many childhood killers, such as polio and measles, in subsequent years. Outreach clinics now attend villages annually and inoculate new-borns against: Polio, DPT (diphtheria, whooping cough and tetanus), BCG (tuberculosis) and measles. Tetanus injections.

90 Chapter 6: Child mortality which are routinely given to all pregnant mothers, may have effected levels of neonatal mortality.

Finally, the villages have suffered a more recent period of irregular rainfall during the late 1990s, which was ongoing at the date of survey. This has resulted in acute intra­ household food shortages, following the failure of the Arfassa early rains necessary for planting. In 1999 five woredas, including Dodota and Hitosa, were determined by the UNDP as food insecure and in need of emergency food relief assistance (Belay 1999).

Prior to the 1990s the primary cause of death in early childhood was shifto, the symptoms of which included communicable diseases, e.g. measles and polio. However, local clinic records indicate that more recently the primary cause of child illness has related to diseases associated with malnutrition. The five major diseases of early childhood among children brought to Amudee clinic for treatment in 2000 were: protein energy malnutrition (or kwashiorkor - nefaso), skin infections (impetigo, scabies), malaria, ‘diseases of childhood’ - shifto, and respiratory tract infections

6.2 Does improved water supply Impact on mortality?

Both the quality and quantity of water supplies may have an impact on levels of morbidity and mortality (Esrey and Habicht 1986). Previous studies have indicated that many of the health problems experienced in rural Ethiopia can be attributed to the polluted nature and short supply of water (Yohannes, Streatfleld et al. 1992). In 1992 a health survey conducted in Hitosa woreda indicated that 60% of the patients attending the government clinics were suffering from water-borne diseases. Diarrhoea and gastro­ enteritis caused by parasitic, bacterial and viral infections are highlighted as being among the most serious health problems affecting young children.

The major diseases associated with water quality are enteric infections caused by drinking water that is contaminated with infectious water-borne bacteria, viruses and parasites (Guerrant, de Souza et al. 1996). The quantity of water used for washing has been linked with skin, eye and certain enteric diseases, including water-washed diarrhoeal diseases (Feachem, Bums et al. 1978). Numerous studies have outlined the health benefits for children associated with improving water supplies, including reduced

91 Chapter 6: Child mortality morbidity levels and higher child survivorship, (Esrey and Habicht 1986; Esrey, Potash et al. 1991; and review in Curtis 2000).

Consequently, the date of the water point installation (March, 1994) is expected to be an important predictor of improved survivorship in any villages benefiting from the new water supply. To test this assumption, multivariate regression analyses are undertaken using birth history data to examine the demographic change associated with the improved water supply scheme, and to identify the main determinants of child mortality.

6.3 Analyses of the main determinants of mortality

To examine the determinants of childhood mortality at the study site, two sets of analyses were performed. Multivariate logistic regression models describe the partial effects of the explanatory variables on: the risk of neonatal (<1 month) death for all births recorded in the dataset; postneonatal (1-12 months) death (excluding births in the last year); and later childhood (13-60 months) deaths (excluding births occurring within 5 years of the survey date). Table 6.2 illustrates the distribution of births, deaths and percentage of children that die, for each of the variables included in the logistic model. The results of the multivariate logistic regressions presented in Table 6.3 are expressed as a ratio of the log-odds of the beta co-efficient.

Additional multivariate event history analyses (or hazards modelling) describe the monthly probabilities of death across the entire childhood period (0-60 months) for a smaller sample which includes only births occurring within 6 years of the survey date recorded in the event history calendar (Tables 6.4, 6.5, 6.6). Other studies have identified a correlation of survival outcomes between sibs within the same household (Das Gupta 1990; Curtis, Diamond et al. 1993). To avoid any large effects associated with death clustering within families the dataset is restricted to two births per woman, producing a total of 1950 births to be included in the analysis.

The covariates entered in the analysis vary slightly across the models. Using event history analysis permits the use of time-varying covariates (e.g. access to water tap stands) [Chapter 2]. Further explanatory variables collected in the event history calendar

92 Chapter 6: Child mortality for the 7 years prior to the survey date are also included in the event history model. These include, season of birth, and husband’s marital status (presence of polygyny). To control for variation between villages, which may not be determined by factors other than altitude (e.g. access to new roads, clinics), the variable village altitude is replaced by a covariate for each of the 7 villages. Covariates used in the logistic regression analyses, but excluded in the event history modelling are economic activity of the head of household and access to toilet, since both afford very small sample sizes. For similar reasons the three categories for maternal education level are more simply grouped as presence/absence of any schooling.

An additional model (Table 6.5) includes variables on the status of previous births (survival status & birth interval lengths) for a reduced sample of those women experiencing at least two births (n= 1225). A final model (Table 6.6) includes a covariate for maternal physical condition (BMI), for the reduced sub-sample of women who had also been measured during the anthropometric survey date (n= 630). Controls for time and time squared (in months) are entered into all the event history models for analyses, since the risk of death is likely to vary as a non-linear function of length of exposure.

In the results section below, explanatory variables have been broadly categorised into year of birth, season of birth, village site, demographic characteristics, social and economic indices, access to water points, preceding birth characteristics and maternal physical condition.

6.3.1 Year of birth

There is no evidence of a secular trend in neonatal deaths across the time period covered in the dataset (Table 6.3). However, the risk of infant and childhood deaths has significantly reduced for the cohort of births bom since the early 1990s coinciding with access to regional vaccination programmes and the construction of government clinics in nearby market towns of Iteya and Amudee in 1994. In the event history analysis (Table 6.4) there is no clear secular trend in risk of child death across the six-year observation period prior to the interview (1994-2000).

93 Chapter 6: Child mortality

6.3.2 Village site

The altitude of the village that a child is bom into has no statistical effect on the probability of experiencing neonatal or infant death; however, the direction of the effect suggests that infants in low-lying villages may be exposed to a greater risk of death (Table 6.3). There is a strongly statistical effect of altitude on the probability of later child death. Children in lowland villages are 75% more likely to die than those residing in highland villages. The effect reaches statistical significance in later childhood, an age group that has started or completed the weaning process. As such, they are no longer receiving the full nutritional and immunological protection of breast milk. Among the Oromo supplementary foods, including porridge (barley, milk and butter), are introduced at around 6 months, and full weaning takes place around the age of two or at the next pregnancy.

Highland areas are associated with colder temperatures and higher annual rainfall [Chapter 3: Table 3.1]; highland villagers complained of a higher prevalence of acute respiratory tract infections and common colds. Low-lying areas experience higher temperatures and lower rainfall and suffered greater prevalence of malaria and chronic malnutrition. These areas are generally less agriculturally fertile than higher areas and soils are unable to support the variety of crops found in higher areas (e.g. vegetables, barley, teff and wheat). Low-lying areas traditionally provided suitable grazing land for cattle, and child diets were supplemented with milk products. However, following land re-distribution in 1980s and population increases, there has become less grazing land available to support the herding economy. Furthermore, more recent crop failures have forced many farmers to sell any remaining cattle. Villagers and local health workers identified malnutrition as the most common cause of early childhood illness across villages.

Village-level variation is explored in the event history analysis (Table 6.4). Children bom in Daya Dodota (a highland village, with closest access to the town) have a reduced risk of dying than offspring bom in the reference village, Terro Moyee (the most remote village in the sample, in a highland area). However, children bom in Bekare Washo (a lowland village) have a significantly higher monthly probability of dying.

94 Chapter 6: Child mortality

6.3.3 Demographic characteristics

The sex of the child is a good predictor of its likelihood of death during each stage of early childhood (Table. 6.3). During the neonatal and infant periods males have an increased probability of dying; while during later childhood males have a survival advantage relative to females. In the event history analysis (Table 6.4), the sex of the child is still a fairly good predictor of monthly probability of death. Female births have a 50% lower risk of dying during early childhood.

The inherent biological weakness of males relative to females can explain the higher risk of early male death and has been well-documented in medical and scientific literature [this is discussed more extensively in Chapter 7]. However, the improved survivorship of males over females in later childhood may indicate a cultural rather than biological proximate mechanism. Among the Oromo there is a cultural preference for sons over daughters [Chapter 7: Section 7.3.]. Cultural preferences may be expressed through behaviours which are likely to improve son’s chances of survival in childhood, e.g. prolonged breast-feeding or preferential allocation of food after weaning. There is no evidence to suggest that sons are breast-fed for a longer time period than daughters (Chapter 7: Section 7.3) and, unfortunately, there are no data available to confirm more detailed breast-finding behaviours or intra-household food allocation practices.

The effect of birth order (or parity) on risk of dying in childhood is not strong. High parity births (+7) have a 50% greater risk of dying in the neonatal period than first births, but not in later age periods (Table 6.3). In the event history model (Table 6.4), parity has a stronger effect on risk of death, with all non-first births experiencing a lower probability of dying in each month following birth. Taken together the results of the two models indicate that both first birth and very high parity births (7+) are exposed to the greatest risk of dying during childhood. First births are inclined to be the most vulnerable since young mothers are both physically immature and inexperienced. A tendency for the highest parity births to die may indicate that having a large family size incurs a large energetic cost, which can not be met either by the mother or through sib- assisted childcare. A parity-specific decline in maternal condition in Oromo women [identified in Chapter 5] may explain the adverse effects of birth-order on neonate survivorship.

95 Chapter 6: Child mortality

Despite being highly correlated with parity, maternal age at birth has an independent and opposing effect on childhood mortality. After controlling for the effect of parity, younger mothers (<20) generally have a higher probability of offspring death than older mothers. Mothers below 30 years of age have an elevated risk of experiencing a neonatal death, and those below 20 years of age have a higher likelihood of infant mortality compared to older mothers. In later childhood the older mothers (30-39 years) also experience improved offspring survivorship relative to the younger age groups. In the event history model, maternal age at birth has no independent statistical effect on monthly probability of dying; however, there remains a tendency for older mothers (30+) to have a higher monthly risk of early child death.

6.3.4 Indicators of household social and economic status

The locality of a newly married couple’s household and the proximity of kin is determined by the form of marriage arrangement undertaken [described in Chapter 3: Section 3.2.3]. Locality is entered into the model as a simple indicator of the proximity of maternal female kin (e.g. grandmothers, unmarried younger sisters), since it may predict risk of mortality. Kin selection theory, or ‘the grandmothering effect’, predicts that childcare behaviour should be biased in favour of genetic kin. The presence of female kin has been associated with improved child survivorship and higher fertility in some subsistence-based populations (Sear 2002; Sear 2000; Hawkes, O’Connell et al 1998; Hawkes, O’Connell et al 1997).

In the regression analysis the locality of the household has no independent significant effect on the probability of death during childhood (Table 6.3). However, in the event history model, the locality of the household is a strong significant predictor of the child’s monthly risk of death (Table 6.4). Matrilocal residence is rare, but those women marrying within their own villages achieve improved offspring survivorship. These women are likely to have improved assistance with childcare, since they may have easy access to their biological female kin (sisters, mothers etc.) than those women living with their affines.

96 Chapter 6: Child mortality

Religion (ethnicity) is a strongly significant predictor of neonatal mortality. Muslim (Arsi Oromo) households are exposed to a 50% increased probability of death in the first month of life, compared with Orthodox (Shoa Oromo) households. However, there is no statistical effect in infancy or later childhood. One possible explanation for the high rates of neonatal death among the Arsi Oromo may relate to the high prevalence of the viral infection Hepatitis B, which occurs in the population. Clinical studies of Arsi Oromo women indicate that up to 80% of the 20-40 year old women are carriers for the virus (Pasquini, Bisanti et al. 1988). Hepatitis B infection during pregnancy has negative consequences for both mother and baby, and has been linked with foetal mortality, especially in the final trimester of pregnancy (Kwast and Stevens 1987). Cultural practices known to occur among the Arsi Oromo, including adult female circumcision and body piercing (Terete 2000), are considered major determinants of seropositivity. However, in the event history analysis neither religion, nor husband’s marital status (polygyny), has a separate effect on child death.

Herd size is an indicator not only of household wealth and social status but also the availability of food. Milk products from cattle (milk and butter) are an important source of nutrients for both children and adults. At all three stages of childhood the presence of cattle is a good predictor of offspring survivorship. Households with at least one cow have an increased probability of their offspring surviving early childhood than households with no cattle in both logistic and event history models. The Kaplan Meier survival plot in Figure 6.2 illustrates that the strongest effect is after weaning in later childhood, with the largest herd size (6+) associated with the lowest monthly probabilities of child death.

Any level of female education is rare in rural Ethiopia, and opportunities for schooling are confined to those women from the highest status households [Chapter 3: Table 3.3]. In the logistic model (Table 6.3), level of maternal education has no clear effect on neonatal mortality; however, it is negatively associated with probability of infant or later childhood death. The monthly probabilities of dying, in relation to the level of maternal education is illustrated in a Kaplan Meier survival plot in Figure 6.3, The positive effect of education on survivorship may only reach statistical significance during infancy due to the small sample sizes for categories of educated women (Table 6.2). Maternal education is likely to be critical for surviving infancy, since at this age

97 Chapter 6: Child mortality supplementary foods are introduced and the infant is exposed to new pathogens. Any form of schooling is associated with a reduced monthly probability of early childhood death in the event history analysis (Table 6.4).

The main subsistence activity of the head of household is an indicator of the household economy. Maize farming is a less valuable crop at the market and is associated with low-income families and a dependence on one crop. Wheat is farmed in the most fertile areas of the highlands, exacts a higher price at the market and is often mixed with other crops, e.g. barley or teff. Tradesmen and government officials are a tiny group consisting of the highest earning households. However, in the analysis, economic activity has no significant independent effect on neonatal or infant mortality rates, but in later childhood the highest earning households have a 50% reduced risk of death than the lowest earners. High earning households may be in a position to buy important foodstuffs, e.g. milk and butter, for their offspring at the market and may be buffered from losses to the cattle herding economy experienced by most villagers.

In each of the three villages with the water points, only four model pit latrines had been built at the date of survey. These were constructed within the homesteads of high status community members, e.g. a Peasant Association Leader, militia member and water tap attendant. For the majority, excreta are disposed in the fields behind the household compound. The results of the regression analysis indicate that household access to a pit latrine is associated with a reduced risk of childhood mortality. However, this may only become statistically significant during later childhood, since the child will be exposed to more pathogens once it has been weaned (Table 6.3). A child with access to a pit latrine is at a 60% reduced probability of dying between the ages of one and five.

6.3.5 Season of birth Seasonal patterns of food availability and disease may also influence early childhood development and mortality risks. Season of birth is a strongly significant predictor of monthly probability of death (Table 6.4). Births occurring during the wettest seasons, which are also the months of lowest food availability (ganna July-Sept), are associated with an increased likelihood of death per month than those bom in harvest and dry

98 Chapter 6: Child mortality season months {birra Oct-Dec and bona Jan-Mar). The monthly survival function is illustrated in a Kaplan-Meier plot in Figure 6.4.

Diseases recognised by the Oromo as being prevalent during the wet season include respiratory infections, dysentery and stomach problems arising from eating unripe grain and from contracting malaria in the low-lying areas. However, overall, the main cause of infant deaths is related to problems of malnutrition associated with the hungry season, commonly termed nefaso\ symptoms include the swelling of face and stomach characterising protein-energy malnutrition (kwashiorkor).

6.3.6 Access to water tap stands Access to village-level water tap stands, entered into the model as a time-varying covariate, has no statistical effect on the probability of dying in early childhood (Table 6.4). Furthermore, an odds ratio effect above 1.00 indicates that, in those villages with access to water development, the timing of the installation of tap is associated with an increase in probability of child death. These results fail to support the prediction that an improved water supply should increase early childhood survivorship, both through improved hygiene opportunities, access to clean drinking water and reduction of women’s workloads during pregnancy.

There are several reasons why the water taps may not have improved child survivorship in these villages. Collecting water from the new supply may not have increased the use of water used for washing/hygiene purposes. Focus group discussions in the villages reveal that in villages with new taps women have not increased the volume of water collected, since women are now charged per container. Chapter 3 (section 3.4) outlines household water use/consumption practices at the study site.

At the point of survey, there was no hygiene or health education programme operating in any of the villages surveyed. Hygiene conditions within the villages were very poor, which may have facilitated the transmission of diseases within and between households. Human excreta and waste were disposed on nearby paths or fields, well within the reach of small children. At night waste was disposed within the household compound, then swept to the edge of the compound with branches in the morning. Animals, such as

99 Chapter 6: Child mortality young calves, goats and chickens continued to be kept in the family hut, where cooking, eating and sleeping took place. Furthermore, in villages within the water project, the tap stand itself had become a social meeting place for women and their children, which may have facilitated disease transmission between families.

The failure of the new water supply to improve child survivorship may be the result of prevailing poor hygiene and low household water consumption. Studies have indicated that the availability and reliability of water supply, irrespective of its quality, is the most important factor in reducing childhood disease levels (Hughes 1983). Moreover, improved sanitation and hygienic behaviour is crucial (Esrey 1996). Overall, the greatest health problems encountered during early childhood in this population are likely to relate to increasing levels of undemutrition.

6.3.7 Previous birth characteristics Numerous studies on the child spacing versus offspring survivorship have revealed that short birth intervals (usually <24months) are associated with higher offspring mortality (De Sweemer 1984; Retherford, Choe et al. 1989; Koenig, Phillips et al. 1990; Pedersen 2000). Reduced offspring survivorship has been attributed to various causes, including: maternal depletion (Tracer 1991); sib competition (Pebley, Hermalin et al. 1991; Aaby, Pison et al. 1995); and early weaning (Palloni and Tienda 1986; Retherford, Choe et al. 1989).

The results of the multivariate event history analysis with the additional explanatory variables for the characteristics of the previous birth for the reduced sample of women who have experienced at least two live births (n=1225) are presented in Table 6.5. After controlling for the partial effects of the covariates observed in Table 6.4, the survival status of the previous child is not a significant indicator of the likelihood of early childhood death (indicating that there is no death clustering between sibs). Categorised length of previous birth interval is also included as an indicator of a woman’s recent birth spacing pattern. After controlling for any family effects, a short preceding interval (<24 months) is associated with a greater monthly probability of dying than longer spaced preceding intervals. Among this energetically stressed population a pattern of short birth-spacing is likely to increase offspring mortality by removing the nutritional

100 Chapter 6: Child mortality and immunological effects of prolonged lactation and increased sib-competition for childcare, particularly in late childhood between ages 2 and 4 (Le Grand and Phillips 1996)

6.3.8 Maternal body condition Table 6.6 demonstrates the results of the final event history model with the additional explanatory variables for maternal anthropometric status, measured as body mass index (BMI). Maternal BMI is a strong predictor of monthly risk of early childhood mortality. Malnourished women, those below the 18.5 limit for chronic energy deficiency, have a two-fold risk of experiencing a child death than better-nourished women. The probabilities of dying in each month following birth for maternal energetic status are illustrated in the Kaplan Meier plot in Figure 6.5. Energy deficient women may be exposed to greater risks of child death, both because they have a tendency to produce vulnerable, low birth-weight babies (Tafari, Naeye et al. 1980), but also because they are unable to sustain the full energetic costs of postnatal lactation and childcare.

101 Chapter 6: Child mortality

Table 6.2 Distribution of births and deaths by each explanatory variable

Variable Neonatal (i Qo) Postneonatal (nPi) Later child ( 43^ 12) births deaths births deaths births deaths Year of birth <1978 1413 124 1289 190 1099 98 1979-80 407 34 373 51 322 38 1981-82 446 37 409 49 360 27 1983-84 507 46 461 42 419 46 1985-86 573 54 519 59 460 42 1987-88 571 50 521 59 462 37 1989-90 562 52 510 67 443 33 1991-92 630 58 572 48 524 39 1993-94 642 45 597 47 550 30 1995-96 699 45 654 51 - - 1997-98 780 41 739 51 -- 1999-00 565 26 ---- Village altitude Highland 4328 306 3733 356 2628 173 Lowland 3467 306 2921 358 2011 217 Sex Male 4027 345 3427 390 2374 180 Female 3768 267 3217 324 2265 210 Parity 1 1438 131 1220 152 849 86 2-3 2373 185 2002 227 1416 132 4-6 2385 157 2079 205 1497 117 7+ 1599 139 1343 130 877 55 Maternal age at birth <20 1924 178 1633 213 1172 116 20-29 3996 297 3419 352 2436 215 30-39 1719 128 1459 135 969 54 40+ 156 9 133 14 62 5 Locality Matrilocal 1057 74 906 90 641 49 Patrilocal 6738 538 5738 624 3998 341 Religion Orthodox 1984 113 1721 157 1235 90 Muslim 5811 499 4923 557 3404 300 Herd size None 1186 106 972 110 645 84 1-5 cows 4976 383 4246 465 2938 245 6+ cows 1633 123 1426 139 1056 61 Maternal education 7206 576 6189 687 4410 374 None 311 25 237 15 128 11 1 -3 years 278 11 218 12 101 5 4+ years Econom ic activity Maize farm 2993 260 2537 295 1758 184 Wheat farm 4471 324 3825 380 2684 196 Trades/Govnt. 331 28 282 39 197 10 Access to toilet Pit latrine 318 16 276 24 198 5 No toilet 7477 596 6368 690 4441 385 Sample 7795 612 6644 714 4639 390

102 Chapter 6: Child mortality

Table 6.3 Multivariate logistic regression of the probability dying in three stages of childhood (the neonatal period, postneonatal and later childhood) [tp=<0.1, **p=<0.05]

Variable Neonatal death Postneonatal Later child death (0-1 months) death (1-12mths) (1-5 years) odds ratio odds ratio odds ratio Year of birth <1978 0.617 1.916** 1.595** 1979-80 0.955 1.813** 2.433** 1981-82 0.979 1.594** 1.463 1983-84 1.045 1.166t 2.296** 1985-86 1.075 1.466t 1.774** 1987-88 1.029 1.4841 1.588t 1989-90 1.085 1.737** 1.441 1991-92 1.084 1.067 1.149 1993-94 1.00 1.00 1.00 1995-96 0.740 0.994 - 1997-98 0.596 0.876 - 1999-00 0.527f -- Village altitude Highland 1.00 1.00 1.00 Lowland 1.112 1.206 1.775** Sex Male 1.00 1.00 1.00 Female 0.815** 0.836** 1.193t Parity 1 1.00 1.00 1.00 2-3 0.939 0.975 0.999 4-6 0.862 0.954 0.826 7+ 1.516* 1.079 0.939 Maternal age at birth <20 1.190 1.247t 0.970 20-29 1.00 1.00 1.00 30-39 0.734f 0.882 0.667** 40+ 0.543t 1.090 1.100 Locality Matrilocal 1.00 1.00 1.00 Patrilocal 1.000 1.041 0.978 Religion Orthodox 1.00 1.00 1.00 Muslim 1.520** 1.139 0.947 Herd size None 1.00 1.00 1.00 1-5 cows 0.780** 0.902 0.624** 6+ cows 0.830 0.812t 0.488** Maternal education None 1.00 1.00 1.00 1-3 Years 1.153 0.640* 0.953 4+ years 0.606 0.531** 0.695 Economic activity Maize farming 1.00 1.00 1.00 Wheat farming 0.958 1.030 0.981 Tradesman/ Go 1.042 1.313 0.456** Access to toilet Pit latrine 0.837 0.945 0.391** No toilet 1.00 1.00 1.00 Intercept -2.54** -2.05** -2.24** Births (n) 7795 6644 4639 Deaths 612 714 390

103 Chapter 6: Child mortality

Table 6.4 Multivariate event history regression of the probabiiity dying in early childhood (sPo) [tp=<0-1, **p=<0.05, ***p=<0.005]______Variable Probability of child death births deaths % dead odds ratio P Year of birth 1994 116 21 18.1 0.629 0.087f 1995 205 38 18.5 1.002 0.641 1996 303 54 17.8 1.00 - 1997 384 45 11.7 0.629 0.059t 1998 377 48 12.7 0.903 0.202 1999 376 38 10.1 0.671 0.418 2000 189 12 6.3 -- Village Daya Debeso (high) 225 22 9.8 0.634 0.212 Daya Dodota (high) 358 38 10.6 0.481 0.052t Terro Moyee (high) 260 31 11.9 1.00 - Hurturbe (high) 232 31 13.4 1.215 0.408 Bekare Washo (low) 292 51 17.5 1.510 0.046** Reissa Michiko (low) 264 33 12.5 0.791 0.310 Debula Saapo (low) 319 50 15.7 0.838 0.539 Sex Male 991 145 14.6 1.00 - Female 959 111 11.6 0.794 0.074t Parity 1 305 43 14.1 1.00 - 2-3 563 65 11.5 0.624 0.058T 4-6 555 66 11.9 0.571 0.046** 7+ 527 82 15.6 0.771 0.437 Maternal age at birth <20 358 44 12.3 0.975 0.919 20-29 951 118 12.4 1.00 - 30-39 556 79 14.2 1.006 0.980 40+ 85 15 17.6 1.094 0.797 Locality Matrilocal 253 18 7.1 0.481 0.004*** Patrilocal 1697 238 14.0 1.00 - Religion Orthodox 500 55 11.0 1.181 0.461 Muslim 1450 201 13.9 1.00 - Herd size None 352 57 16.2 1.00 - 1 -5 cows 1252 156 12.5 0.652 0.007*** 6+ cows 346 43 12.4 0.667 0.067t Maternal education No formal education 1653 231 14.0 1.00 - 1 + years 297 25 8.4 0.669 0.072t Husband’s marital status Polygynous 447 59 13.2 0.871 0.400

Monogamous 1503 197 13.1 1.00 - Season of birth

Dry season {bona) 744 75 10.1 1.00 - Early rains {afrassa) 578 84 14.5 1.312 0.099T Long rains (ganna) 623 97 15.6 1.472 0.014** Access to water With taps (Time-vary) 446 46 10.3 1.238 0.442 Time (months) 0.842 0.000*** Time^ 1.002 0.000*** Intercept -2.373 0.000*** Births (n) 1950 Child deaths 256

104 Chapter 6: Child mortality

Table 6.5 Multivariate hazard regression of the probability dying in early childhood (sqo) with preceding birth covariates (other variables as Table 6.4, not shown)

Variable ______Probability of child death births deaths % odds ratio P dead Survival status of previous child Alive 1031 112 10.9 1.00 Dead before next birth 194 20 10.3 1.145 0.559 Length of preceding interval 289 49 17.0 1.750 0.006** <23 months 886 90 10.2 1.00 24-47 months 50 3 6.0 0.651 0.475 48+ months Intercept -2.561 0.00*** Births (n) 1225 Child deaths 142 Notes: tp=<0.1, **p=<0.05, ***p==<0.005

Table 6.6 Multivariate hazard regression of the probability dying in early childhood (sqo) v anthropometric covariates (other variables as Table 6.4, not shown)

Variable Probability of child death Births deaths % dead odds P ratio Body mass index <18.5 120 23 19.2 1.815 0.038** 18.5-21.5 345 40 11.6 1.00 22.5+ 165 14 8.5 0.711 0.295 Intercept -1.644 0.093t Births (n) 630 Child deaths 77 Notes: fp=<0.1, **p=<0.05, ***p= <0.005

6.4 Concluding remarks

The first month of life remains the most dangerous period of life for a young Oromo child. Despite the improvements to health care facilities, which have reduced later child mortality, levels of neonatal mortality in these villages have remained high across the 20-year observation period. The robusticity of this figure relates, in part, to the high prevalence of low birth weights (<2500g) in villages; over 65% of the deliveries bom at the Amudee clinic are underweight (<2500g) (Amudee clinic nurse Pers. Comm.). Furthermore, mothers rarely seek prenatal or antenatal care and are brought to the clinic for the delivery only if there are severe complications, e.g. obstmcted delivery.

105 Chapter 6: Child mortality

An association between high rates of neonatal death with both low birth weights (<2500g) and pre-term deliveries is consistent across energetically stressed populations. In Bangladesh, low birth weights approximately doubles the risk of neonatal death (Yasmin, Osrin et al. 2001) and preterm delivery was implicated in three-quarters of neonatal deaths. High maternal workloads are compounded by inadequate nutrient intake during pregnancy and close-birth-spacing patterns, placing energetic stresses on mothers and their offspring in utero. Although maternal physiology has been designed to favour offspring survival beyond the early trimesters of pregnancy, e.g. by lowering basal metabolic rates in the Gambia (Prentice, Whitehead et al. 1981), under extreme conditions of ecological adversity an increase in maternal energetic stress is associated with lower infant birth weight and pre-term delivery. A low caloric intake during the Dutch Hunger Winter of 1944-45 increased the frequency of pre-term births (Kline, Stein et al. 1989) and conversely, the introduction food supplements for pregnant women in rural Gambian villages decreased the frequency of low birth weight babies (Prentice, Whitehead et al. 1981).

As stated previously, infant and later child death has declined considerably since the introduction of immunisation programmes, which may have reduced the prevalence of a number of fatal early childhood diseases, e.g. polio and measles. However, a quarter of reproductive-aged women still reported that they had not immunised any of their children [Chapter 5: Table 5.3]. The under five death rate in this rural population remains extremely high (232.3/1000 live births). Evidence presented here indicates that at the survey date the introduction of new village-level piped-water tap stands alone has not contributed towards the improvement in child survivorship. Furthermore, the analyses indicate that both mothers [Chapter 5] and their young children may be vulnerable to shorter birth-spacing patterns and larger family sizes, as well as the nutritional threat posed by recent food shortages.

106 1.0 10 Fig 6.2 Fig 6.3 K-M plot of the K-M plot of the effect of herd size effect of .9 on survival maternal educ­ ation on survival

Herd size Maternal education 8 6+ cattle 4+ years

8 1-5 cattle (/) 7 w 1-3 years c o c -2 o o None o Q. Q. none 2 o o_ 6 .7 -10 0 10 20 30 40 50 60 -10 0 10 20 30 40 50 60

Months since birth Months since birth

1 0 1.0 Fig 6.5 Fig 6.4 K-M plot of K-M plot of the effect of the effect of maternal BMI season of on survival birth on ) survival .9 Maternal BMI Season of birth 21,5+ long wet (june-sept) .8 .8 18.5-21 49 short wet (mar-may) (A C (A o C ■■C o o ■•C Q. <18.5 o CL dry (oct-feb) 2 7 o CL .7 -10 0 10 20 30 40 50 60 -10 0 10 20 30 40 50 60 7

Evidence for adaptive sex-biased patentai investment

‘Abban durbaa maqaa hinqabu /The father of a girl has no name’ [traditional Oromo saying]

7 .1 Sex biased investment theory In the early part of the 20^ Century (Fisher 1930) identified that an overall parental investment bias favouring one sex within a population would be maladaptive. The stabilising effect of natural selection at the individual-level should maintain the population sex-ratio equilibrium. More recently Trivers, Willard and Maynard-Smith, among others, have argued that, under certain well-defined conditions, a strategy in which individuals within a population bias their investment in the sexes differentially could be adaptive (Trivers 1972; Trivers and Willard 1973; Maynard Smith 1980).

Trivers-Willard’s model of sex-biased parental investment (Trivers and Willard 1973) states that the physical condition of the mother will influence her pattern of investment in offspring in an adaptive manner. The central tenet of this theory is that a mother’s investment influences the condition of her young into adulthood and, if the reproductive success of the sexes is differentially influenced by condition, mothers are predicted to bias their investment in favour of the sex of offspring that gives greatest fitness returns per unit of investment, in terms of number of grandchildren.

In polygynous species, such as humans, under favourable conditions parents are expected to bias investment towards male offspring, while under poor conditions they should bias towards female offspring. This interpretation is based on the assumption that traits such as increased body size in adulthood are more important in affecting male Chapter 7: Sex biased investment reproductive success and that males have greater variance in reproductive success. Accordingly, a reproductively successful male could make a greater contribution to parental inclusive fitness than a reproductively successful female; however, under poor conditions a male could be relatively disadvantaged. Parents should manipulate the sex- ratio of offspring in response to internal or environmental cues of resource availability or predictability in order to maximise their own reproductive success.

This argument can be extended across the whole period of parental investment from conception through to early childhood. Parents may adjust the sex-ratio of their offspring either, before birth, through differential foetal loss (biological processes), or post-natally, by biasing the allocation of childcare (behavioural practices). In this chapter I shall examine the evidence for sex biased parental investment before birth and sex biased mortality in early life according to three measures of resource availability:

maternal body condition

season of birth

year of birth

7.2 The relative costs of the sexes The rationale for adaptive sex-ratio manipulation influenced by resource availability, relies on the relative costs and survival of the sexes. Clinical studies have demonstrated that male babies are larger and more energetically expensive for the mother to produce (Copper, Goldenberg et al. 1993), but are also more vulnerable than females in infancy. Male children are exposed to a greater risk of intra-uterine mortality (Kellokumpulehtinen and Pelliniemi 1984; Jakobovits, Jakobovits et al. 1987), pre-term delivery (Cooperstock and Campbell 1996) and malnutrition, infectious disease and subsequent mortality persisting in early childhood (Read, Troendle et al. 1997; Morris, Victoria et al. 1998) than females.

Furthermore, traits such as small body size are more important in affecting male reproductive success than female reproductive success. Poor intra-uterine conditions are associated with low birth weight-babies (Thomson 1959; Stein, Susser et al. 1975), stunted child growth (Eveleth and Tanner 1990), with early insults having adverse effects on later development (Lucas, Morley et al. 1998), including small body size in adulthood (Golden 1994). Male offspring are more adversely affected by low birth

109 Chapter 7: Sex biased investment weights than female offspring, e.g. suffering lower IQ levels (Matte, Bresnahan et al. 2001). Furthermore, in terms of fitness, males of short stature are relatively disadvantaged later in life in finding mates (Phillips, Handelsman et al. 2001), experiencing lower reproductive success in a number of traditional and developed populations (Pawlowski, Dunbar et al 2000; Winkler and Kirchengast 1994; Hill and Hurtado 1996). Investing in low birth-weight male offspring, under conditions of poor resource availability would clearly incur fitness costs; a more successful strategy would be to invest in daughters.

7.2.1 Testing adaptive models of sex-biased investment Experimental and observational animal studies have demonstrated a facultative adjustment in sex-ratio at birth in response to maternal condition, e.g. in gulls (Nager, Monaghan et al. 1999), red deer (Clutton-Brock and lason 1986) and mice (Rivers and Crawford 1974); however, there is no clear supporting evidence from a human population. A positive effect of stature on proportion of male births has been identified among male San hunter-gatherers (Winkler and Kirchengast 1994) and urban women in Central African Republic (Andersson and Bergstrom 1998), however, the effects are weak, being constrained by small sample sizes. In industrialised populations, maternal socio-economic conditions have been associated with differential foetal mortality, i.e. sex-ratio at birth (Chacon-Puignau and Jaffe 1996). However, to my knowledge no previous studies have identified a strong relationship between a direct measure of current resource availability (maternal body condition) and sex-ratio at birth within one population.

Evidence for post-natal investment practices being influenced by maternal condition has been demonstrated in both traditional and industrialised populations. These studies indicate that well provisioned mothers bias their child care towards male offspring leading to higher female mortality, while poorly provisioned mothers favour daughters and experience higher male mortality (Dickemann 1979; Boone 1988; Cronk 1989; Abemethy and Yip 1990; Gaulin and Robbins 1991). However, an equal number of of later studies have only found negative, weak or inconclusive evidence of a Trivers- Willard effect (Maconochie and Roman 1997; Keller, Nesse et al. 2001; Koziel and Ulijaszek 2001). Later studies may not have found adaptive sex ratio manipulation since

110 Chapter 7: Sex biased investment these have concentrated on larger sample census data, which may have included a number of sub-populations responding to differing local conditions (see p 128).

Factors other than maternal condition may also influence the costs and benefits of investing in sons and daughters. These include the level of competition for local resources (Johnson 1988) and the existing family composition (Muhuri and Preston 1991; Madise and Diamond 1995) and the level of help provided by the sexes, either in rearing younger offspring (Emlen, Emlen et al. 1986), or by contributing to the household economy (Caine 1977). Whether differential biases in sons and daughters represent adaptive patterns of investment in humans remains highly disputed.

7.3. C ultural sex preferences ‘In Arsi, women performing the ritual of obaaxxa dhlquu, washing the m other and child [after a birth], make a clear gender distinction. The bathing takes place In 4 days for a girl and 5 days for a boy and the ululation Is uttered only 4 times for a girls, but 5 times for a b o y ’ (Terefe 2000).

The Oromo express a cultural preference for sons over daughters, the ideal family being composed of twice as many sons as daughters. The reasons for this preference relate to the system ofgossalcXan exogamy, whereby daughters are given away at marriage and become associated with their husband’s gossa. A daughter’s sons are also less preferred than a son’s sons, being termed 'nhyaatu galuu' (‘he returns to his home after eating’), reflecting an unbalanced relationship in which they take from, but do not contribute to, the household.

When asked directly about distributional biases between children of differing sexes, women tended to deny unequal treatment; however, previous studies have indicated that it is not always safe to assume that intention and behaviour are analogous (Cronk 1991). Among the sample there is some evidence of preferential childcare towards male children with regards to health seeking behaviour, educational opportunity and breast­ feeding [Appendix I: Illustration 12]. Sick sons are more likely to be taken to the clinic than sick daughters (50%/34%: chi square 4.31 p=0.03). Although, overall, there is little access to education, slightly more sons attend schools than daughters (3.3%/2.8%: chi square NS). There are no sex differences in duration of breast-feeding; however, mothers return to menses sooner following the birth of a daughter, which may indicate

111 Chapter 7; Sex biased investment

either a more intensive regime of breast-feeding for sons or that sons feed more. In Chapter 6 (Table 6.3) there is some evidence indicating that females may be exposed to higher risk of death in late childhood (age 1-5 years).

7.3.1 Reported sex-ratio biases The overall sex-ratio of all the births in the dataset has a ‘normal’ mammalian sex-ratio at birth (1.07), indicating that women are not selectively under-reporting either sex. Hazards analysis in Chapter 6 demonstrates that, despite childcare practices favouring sons, in the entire sample of births in the dataset, male new-boms are exposed to a higher monthly probability of dying than female births during childhood (sqo) (Chapter 6: Table 6.4). To examine whether the sex of the previous birth has any impact on the subsequent birth a multivariate logistic regression is performed to assess the probability of the index birth dying in childhood by sex of the previous child (Table 7.1). All non- first births borne five years prior to the survey date are categorised according to the sex of the sib pairs (male-male, male-female, female-male, female-female). A separate model (not presented here), exploring interactions between all the covariates included no significant effects.

Table 7.1 demonstrates that, after controlling for the significant effects of year of birth, parity and mother’s age at birth, the survival status of the previous child is a significant predictor of the next child’s survival. Additionally, there is evidence that having a same sex sib significantly increases the next child’s risk of dying, although the result does not quite reach statistical significance (p=<0.05). The male-male sib combination is exposed the greatest risk of death (Fig 7.1). This may be due to the extra resources required to bear a male offspring, which are not replenished in time to successfully raise a subsequent male birth. The high death rates for female-female sib pair may relate to the cultural preference for sons and the presence of daughter neglect or infanticide. Overall the findings indicates that male offspring are more energetically costly for the mother to produce and more vulnerable than female offspring; however, the cultural preferences relating to ideal family composition may also be influencing childcare behaviour and mortality patterns.

112 Chapter 7; Sex biased investment

Fig 7.1 Proportion of births dying in childhood (<60months) by sex of previous sib

28

CO 2 6

(U 24

male-male female-male female-female male-female

sib combination

Table 7.1 Multivariate logistic regression modelling probability of dying under five by sex of previous sib combination

Variable Probability of death under 5 Odds ratio P Year of birth 1979-80 1.273 0.129 1981-82 1.093 0.572 1983-84 1.112 0.468 1985-86 1.124 0.427 1987-88 1.063 0.676 1989-90 1.233 0.150 1991-92 -- 1993-94 0.865 0.425 Parity (continuous) 1.006 0.808 Maternal age at birth <20 1.356 0.024** 20-29 - - 30-39 0.817 0.105 40+ 0.961 0.905 Survival status of previous sib Dead 1.674 0.000*** Alive -- Sib-sib combination Male - male 1.215 0.083 Male-female 1.055 0.648 Female - female 1.186 0,143

Female - male -- Intercept -1.429 0.000*** Births (n) 3398 Note: survival status of previous sib*sib-sib combination interactions all NS

113 Chapter 7: Sex biased investment

7.4 Does maternal condition influence sex ratios?

7.4.1 Sex ratio at birth Using the anthropometric data collected from the sample of Oromo women it is possible to examine whether maternal physical condition predicts the sex of her most recent birth. Maternal nutritional status is measured both by mid-upper arm circumference (MUAC in cm), which is a simple indicator of peripheral tissue stores of fat and protein, and body mass index (BMI in kg/m^) which is a measure of body fat. Although the energetic status of each woman is unlikely to have remained constant over the observation period (since the birth of the child), this current status measure may be used as an overall indicator of the variation between women at birth.

Separate multivariate regression analyses are performed to identify an association between mother’s current MUAC, BMI and the sex of her last child, controlling for the independent effects of mother’s age, parity and year of birth (Table 7.2). Entered as continuous covariates, both maternal MUAC and BMI are significant predictors of the sex of the child. Within this food-stressed population there is a positive association between sex-ratio at birth and maternal nutritional status. Figures 7.2 (a) and (b) indicate that as nutritional levels decline mothers produce a lower ratio of males to females, supporting the Trivers-Willard prediction.

Fig 7.2 (a) Sex ratio of most recent birth for maternal mid-upper arm circumference

1,6 1.4

1.2 r 1 0.8 n = 7 6 re o 0.6 1=56 2 n=64 0.4 1=62 reX (A 0.2 n = 80 0 <=23 23.1-24 24.1-25 25.1-26 26.1 +

Maternal mid upper arm circumference (cm)

114 Chapter 7; Sex biased investment

Fig 7.2 (b) Sex ratio of most recent birth for maternal body mass index

2 1.8

I: 1.2

cc 1 n=37 o 0.8 2 0.6 n=49 S 0.4 n=63 n=56 n=65 n=69 " 0.2 f 0 <18.5 18.5-19.49 19.5-20.49 20.5-21.49 21.5-22.49 22.5+

BMI (kg/m2)

Table 7.2 Multivariate logistic regression modelling the probability of having a male last birth for two measures of maternal conditionna) maternal mid-upper arm circumference, (b) body mass index (n=424)

Variable Sex at birth (male) Odds ratio P Odds ratio P Year of birth Year of birth 1994 1.067 0.904 1994 1.023 0.966 1995 1.552 0.640 1995 1.515 0.655 1996 - - 1996 -- 1997 0.945 0.926 1997 0.947 0.928 1998 0.998 0.996 1998 0.939 0.908 1999 0.752 0.599 1999 0.774 0.633 2000 0.822 0.723 2000 0.789 0.675

Parity Parity

First birth -- First birth -- Later birth 0.581 0.198 Later birth 0.524 0.524

Maternal age Maternal age at at birth birth <20 0.590 0.139 <20 0.539 0.078t 20-29 -- 20-29 -- 30-39 0.732 0.204 30-39 0.768 0.768 40+ 0.706 0.439 40+ 0.722 0.722

Maternal 1.171 0.002*** Maternal 1.133 0.018** MUAC BMI Intercept -2.619 0.092t Intercept -1.056 0.448 coefficient coefficient Notes: Average time between birth and measurement = 21.6mths Sample: Males n=199, females n=225; |p=<0.1, **p=<0.05, ***p=<0.005 Interaction terms; muac*muac, bmi*bmi; matage*muac, matage*bmi; parity*muac; parity*bmi all non significant

115 Chapter 7: Sex biased investment

7.4.2 Sex biased child mortality To test for differential post-natal investment practices in the same sample of women, analyses are performed to identify sex biases in early childhood mortality. Since few of the women’s most recent births had died prior to the survey date, the dataset is expanded to include each woman’s two most recent births within five years of the survey date, yielding a total of 690 births for the analysis. Multivariate discrete-time hazards regression analysis are used to test the monthly probability of each infant dying in the first two years of life, controlling for maternal age at birth, parity and year of birth.

Entered as a continuous variable maternal MUAC has an inverse relationship with the infant’s monthly risk of dying in the first two years of life (Table 7.3); a similar though non-statistically significant effect is observed for maternal BMI. Fig 7.3 illustrates the higher proportion of child deaths occurring among poorly nourished mothers. However, the sex of the baby is not a significant predictor of its risk of dying. Separate models were run to test for interactions between all the combinations of covariates; however, these were without statistical significance (sex*MUAC, MUAC*MUAC, MUAC*mother’s age, sex*time, MUAC*Year of birth). Notably, there is no evidence of an interaction effect between sex of the child and maternal status on risk of death, suggesting that mothers are not adopting differing strategies to bias their post-natal child care. Sex-biases in mortality may not be observed in this sample, either due to compensatory cultural preferences and male-biased childcare practices, and/or due to biological intra-uterine selection processes allowing only the robust males to survive pregnancy. If selection for robust, virile males has already occurred in utero, and male offspring are disproportionately attributed to healthy, well-conditioned mothers then post-natal male-biased mortality should not be apparent.

116 Chapter 7; Sex biased investment

Table 7.3 Multivariate logistic hazards regression modelling probability of the last two births dying in early childhood (<24 mths) for (a) maternal mid-upper arm circumference, (b) body mass index

Variable Proba bility of death (<2 years) Odds ratio P Odds ratio P Sex Sex Male 1.038 0.883 Male 1.018 0.941

Female - - Female --

Year of birth Year of birth 1994 1.201 0.758 1994 1.125 0.843 1995 1.415 0.463 1995 1.368 0.506 1996 - - 1996 - - 1997 1.157 0.708 1997 1.200 0.636 1998 1.004 0.991 1998 0.978 0.958 1999 0.816 0.648 1999 0.896 0.801 2000 0.903 0.864 2000 1.004 0.995

Parity Parity First birth First birth Later births 0.593 0.289 Later births 0.579 0.259

Maternal age at Maternal age at birth birth <20 0.699 0.455 <20 0.701 0.451 20-29 - - 20-29 - - 30-39 1.277 0.394 30-39 1.346 0.296 40+ 1.191 0.779 40+ 1.748 1.346

Maternal MUAC 0.869 0.038** Maternal BMI 0.177 0.914

Months since -0.320 0.000*** Months since 0.729 0.000*** birth birth Month squared 0.009 0.007** Month squared 1.009 0.004*** Intercept 0.347 0.845 Intercept -0.939 0.568 Births (n) 690 Births (n) 690 Deaths (events) 67 Deaths events) 67 Notes: **p=<0.05, ***p=<0.005 Interaction terms: Sex*Muac; Sex* BMI; Muac*Muac; BMI*BMI; Sex*Time; Muac*Year of birth; BMI*Year of birth; Parity*Mother’s age: ALL non significant

n=146 14

Fig 7.3 n=137 12 Proportion of n=114 n=134 male and £V) 10 n=96 female births dying in early n=72 j □ males childhood V (<24mths) for ■O □ females 0)CO maternal BMI TJ

<18.5 18.5- 19.5- 20.5- 21.5- 22.5+ 19.49 20.49 21.49 22.49

Maternal BMI

117 Chapter 7: Sex biased investment

7.5 Do seasonal patterns Influence sex ratios?

* during the time when the crops are ripe and women’s bodies are happy and reiaxed...then they can conceive and it is around the next rainy season that many births o c c u r .’ [Local informant]

7.5.1 Seasonal foetal loss Seasonal variation in births in non-contracepting populations relate to annual cycles of food availability, disease levels and workloads, which influence both conception frequencies and pregnancy wastage (Leslie and Fry 1989; Bailey, Jenike et al. 1992; Leslie, Campbell et al. 1993; Ulijaszek 1993). Seasonal diseases, such as parasitic infections (including malaria and schistosomiasis), are associated with varying degrees of foetal loss (McFalls and McFalls 1984). The role of maternal nutrition in determining intra-uterine mortality is less well understood (Bongaarts and Cain 1982); however, extremely low levels of food availability are associated with increasing the risk of spontaneous abortion in the early trimesters of pregnancy (Ford, Huffman et al. 1989; Leslie, Campbell et al. 1993). During these stressful seasons male foetuses are more likely to be miscarried than female foetuses (Nonaka, Desjardins et al. 1998).

In Arsi the hungry season occur during the wet season, when food stores are depleted prior to the current year’s harvest. The food shortage is compounded by the need for hard agricultural labour in the fields, and by both maternal and infant diseases. The hungry season may create a marked negative energy balance for adults, including pregnant women who continue to work throughout their pregnancies. A nutritional study undertaken in a neighbouring region of Arsi provides some evidence of reduced nutritional intake during late dry and wet season (but not of negative energy balance) (Fig 7.4) (Ferro-Luzzi 1990). Reduced food intake has been associated with low child growth velocity (Branca 1993), and with the greatest energy stresses occurring in poorest households (Pastore, Branca et al. 1993). In The Gambia a seasonal reduction in weight gain for pregnant women is associated with severe intra-uterine growth restriction. The prevalence of low birth weight babies (<2500g) exceeds 25% during the hungry months (Ceesay, Prentice et al. 1997).

118 Chapter 7; Sex biased investment

Fig 7.4 Seasonal variation in energy Intake (TEI) and expenditure (TEE) for 22 Southern Ethiopian women (redrawn from (Ferro-Luzzi 1990))

2300 Ganna Biraa Bonaa Arfassa 2200 ^ 2100 5 m 2000 TEI c 1900 TEE I 1800 1700 1600 jun aug sep nov dec jan mar may july Month 1986-87

7.5.2 Sex ratio at birth To identify seasonal fluctuations in sex-ratios at birth in the Oromo sample, multivariate logistic regression analyses are performed to examine the probability of experiencing a male birth (Table 7.4). The data collated using the events calendar provided information on the month of each birth over the six years preceding the interview, producing a total of 2492 births for the analysis. However, any analysis of sex-ratio at birth is complicated by the modification of nutrition and growth requirements of the foetus throughout the three stages of gestation and early life. A cycle of nine month gestation and mono-modal rainfall patterns allows for multiple combinations of interactions between the environment and the mother-foetus, so invariably one stage of gestation or breast-feeding will occur under unfavourable conditions. Figure 7.5 illustrates the seasonal shift in sex-ratio at birth across the year. The table below the figure indicates the food availability and expected growth rates at each trimester of pregnancy for each birth month.

The effects of pre- and post-natal exposure to malnutrition may have different outcomes for both child development and maternal condition (Barker 1998). A high growth trajectory established in early gestation, when foetus requirements are small, leads to an increased demand for nutrients in later gestation. These rapid growing foetuses are more vulnerable to under-nutrition than those with a slow growth trajectory. In the Oromo sample, harvest season conceptions, which are born in the wet season (July-Dee), have

119 Chapter 7: Sex biased investment very low sex-ratios at birth, indicating that fewer rapid-growing male foetuses are being successfully brought to term (Fig 7.5). Those births starting on a slower growth trajectory, conceived during the hungry season and born in the dry season (Jan-June), have a male-biased sex-ratio at birth. The statistical analysis in Table 7.4 indicates that, after controlling for the effects of year of birth, parity and mother’s age at birth, this effect is not statistically significant, but there is a tendency for harvest births to be female biased relative to later dry season births.

Fig 7.5 Sex ratio at birth by season of birth

Ganna Biraa Bona Arfassa

Month of birth July-Aug-Sep Oct-Nov-Dec Jan-Feb-Mar Apr-May-Jun Month of conception Oct-Nov-Dec Jan-Feb-Mar Apr-May-Jun Jul-Aug-Sep Trimester 1 high food high food low food low food Trimester 2 high food low food low food high food

Trimester 3 low food low food high food high food Growth rate fast-slow fast-slow slow-fast slow-fast

re o '.w 2 0>X n=567 n=588 n=563 n=774 0.95

0.9 jul-aug-sep oct-nov-dec jan-feb-mar apr-may-jun Season of birth

120 Chapter 7; Sex biased investment

Table 7.4 Multivariate logistic regression model of the probability of having a male birth by season of birth for all births since 1994

Variable Sex at birth (male) Odds ratio P Year of birth 1994 1.195 0.209 1995 1.263 0.124 1996 - - 1997 1.065 0.666 1998 1.064 0.671 1999 0.859 0.307 2000 1.223 0.273

Parity First birth -- Later births 0.615 0.008**

Maternal age at birth <20 0.710 0.012** 20-29 -- 30-39 1.012 0.902 40+ 0.757 0.212

Season of birth Jul-Aug-Sep 0.306 0.884 Oct-Nov-Dec 0.234 0.867 Jan-Feb-Mar - - Apr-May-Jun 0.596 0.942

Intercept 0.532 0.005** Births (n) 2492 Notes; **p=<0.05

7.5.3 Seasonal disease patterns Seasonal patterns of food availability and disease may also influence early childhood development and mortality risks. Diseases recognised by the Oromo as being prevalent during the wet season are respiratory infections, dysentery and stomach problems arising from eating unripe grain and from contracting malaria in the low-lying areas. Since the introduction of regional vaccination programme (DPT/BCG/Measles/Tetanus) in the early 1990s [described Chapter 6], the prevalence of early childhood communicable diseases and overall mortality has started to decline. In subsequent years the main cause of infant deaths has increasingly been related to problems of malnutrition associated with the hungry season, commonly termed nefaso\ symptoms include the swelling of face and stomach characterising protein-energy malnutrition (kwashiorkor).

121 Chapter 7; Sex biased investment

Furthermore, the season of the year in which a child is born has been demonstrated to affect its resistance to infection and post-natal development (Moore, Cole et al. 1999). In what has been termed ‘the foetal origins of disease’ hypothesis, adult disease patterns are programmed by environmental insults in prenatal or early post-natal period (Barker 1998). Hungry season births suffer a sharp escalating death rate from infectious diseases after puberty, which, while the exact nature of the exposure is not clear, may be mediated by intrauterine growth restriction (Moore, Cole et al. 1997).

7.5.4 Sex biased mortality Analyses are performed on the Oromo sample (n= 2492) to identify any seasonal patterns of early childhood mortality. Figure 7.6 illustrates a trend for male births in July to be exposed to a greater monthly risk of dying across the first two years of life than female births and other male births across the year. Multivariate discrete-time hazards regression analysis is used to test the monthly probability of each infant dying in the first two years of life by sex and season of birth, controlling for mother’s age at birth, parity and year of birth. Interaction terms between sex of the child and season of birth are also included. Table 7.5 demonstrates that there is no statistically significant effect of season on risk of dying for each month following birth and no interaction effect between sex and mortality risk.

Fig 7.6 Proportion of births dying in early childhood (<24 months) for month of birth by sex

oct

sep nov

aug ^ dec •% female infants dying

'% male infants dying jul - jan p-=<0.05

jun feb

may mar apr A r f a s s a

122 Chapter 7: Sex biased investment

The findings presented here, while not all reaching statistical significance (p=<0.05), suggest that male offspring conceived during the harvest, start a fast growth trajectory, but may be exposed to higher risks of intra-uterine mortality during later trimesters of pregnancy which coincide with the hungry months. Furthermore, those males that are successfully brought to term in the hungry season may be exposed to a higher risk of early childhood mortality. This pattern of sex-ratio at birth biased against males during the harvest season may have an adaptive explanation. In the Gambia children conceived during the season of plenty, who spent the second trimester with adequate nutrition, experience the lowest birth weights (Moore, Cole et al. 1997). Small body size may not only render males more vulnerable than females to malnutrition and disease following birth, but short stature may reduce male reproductive fitness in later life.

Table 7.5 Multivariate logistic regression modelling probability of dying in early childhood (<24 months) by season of birth

Variable ______Probability of death under 2 Odds ratio P Sex Male 1.094 0.732 Female -- Year of birth 1994 0.862 0.486 1995 0.862 0.520 1996 -- 1997 0.614 0.040** 1998 0.483 0.004*** 1999 0.377 0.000*** 2000 0.143 0.000*** Parity First birth -- Later births 0.778 0.318 Maternal age at birth <20 0.621 0.061t 20-29 -- 30-39 1.025 0.882 40+ 1.512 0.214 Season of birth Jul-Aug-Sep 1.151 0.623 Oct-Nov-Dec 0.779 0.431 Jan-Feb-Mar 1.185 0.561

Apr-May-Jun - - Sex* Season of birth Sex* Jul-Aug-Sept 1.668 0.174 Sex* Oct-Nov-Dec 1.135 0.768 Sex* Jan-Feb-Mar 1.185 0.727 Sex* Apr-May-Jun -- Intercept -1.779 0.000*** Births (n) 2492 Deaths (events) 227 tp=<0.1, **p=<0.05, ***p=<0.005 Interactions term: parity*mother’s age is non-signif.

123 Chapter 7: Sex biased investment

7.6 Are there long-term fluctuations in sex ratios?

During the 20-year period prior to the survey date, rural Arsi experienced ecological, social-political events and demographic changes which altered both village and household-level structure and the subsistence economy (Appendix II). During this time period there were two notable periods of environmental uncertainty: during the 1984-85 nation-wide famine and in the late 1990s, a period of rain shortage and crop failure (reflected in wheat producer prices in Fig 7.7). If parental sex biases respond to changing patterns of resource availability in the way predicted by Trivers-Willard, then sex-ratio at birth should also vary across this time period.

Fig 7.7 Cereal producer prices {Ethiopian Birr per metric ton) since 1975 [(1975-1994) national data compiled from (PAO 2001) (1998-1999) Arsi data from UNDP-EUE report, (Belay 1999)]

_ 2500 5 2000

.« 1500

1000

•o 500

1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 Year

Fig 7.8 Trends in sex ratio at birth (1979-2000)

I o 1 s (/} 0.7 0.6 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999

Year of birth

124 Chapter 7; Sex biased investment

7.6.1 Sex ratio at birth Figure 7.8 demonstrates the trends in sex-ratio at birth. The graph indicates that there are two periods of recent history in which the sex-ratio of the population was biased towards female births. Between 1981 and 1986, coinciding with the nation-wide famine of 1984-85, and among the 1999-2000 cohort, following more recent food shortages. However, multivariate regression analyses indicate no statistical association between year of birth (entered into the model as groups of 2 years) and the sex at birth, controlling for the independent effects of mother’s age and parity (Table 7.6).

Table 7.6 Multivariate logistic regression modelling probability of having a male birth by birth year cohort (1979-2000)

Variable Sex at birth (maie) Odds ratio P Year of birth 1979-80 1.175 0.208 1981-82 0.960 0.742 1983-84 0.966 0.770 1985-86 0.953 0.675 1987-88 1.062 0.606 1989-90 1.096 0.433 1991-92 -- 1993-94 1.108 0.359 1995-96 1.052 0.646 1997-98 0.998 0.987 1999-00 0.888 0.306

Parity

First birth - - Later births 0.739 0.001***

Maternal age at birth <20 0.855 0.060t 20-29 - - 30-39 1.001 0.950 40+ 0.831 0.264 intercept 0.332 0.006** Births (n) 6382 Notes: fp=<0.1, **p=<0.05, ***p=<0.005

7.6.2 Sex biased child mortality An examination of infant mortality rates over time indicates that there has been an over­ all trend towards declining mortality rates over the 30 years preceding the survey date; the sharpest drop coinciding with the introduction of vaccination programmes in the early 1990s (Fig 7.9). However, through the mid-1980s the rate of decline levelled out, possibly due to the effects of the 1984-85 Ethiopian famine. To examine the relative survival of the sexes across this time period, multivariate logistic regression analyses

125 Chapter 7: Sex biased investment are undertaken for the entire sample of births in the data set (n= 6382). Table 7.7 demonstrates that, while there has been a general reduction in the risk of dying below the age of 2 across the observation period, only those births occurring since 1993 have a statistically reduced mortality risk. Males bom during the famine years, 1983-86 and 1995-2000 are exposed to a greater risk of early childhood death than females. Figure 7.10 illustrates the surplus of male child deaths occurring during the periods of food shortage. Furthermore, Figure 7.11 demonstrates that it is the large fluctuations in male neonatal deaths (deaths within the first month of life), which explain these childhood mortality patterns. Low birth weights, pre-term delivery (Yasmin, Osrin et al. 2001) as well as breast-feeding practices (Huffman, Zehner et al. 2001) may be responsible for these early neonatal deaths.

These results indicate that the sex-ratio at birth and post-natal mortality fluctuates according to the environmental conditions over time. During times of severe resource restriction young male offspring may be relatively disadvantaged, suffering higher levels of foetal loss during pregnancy and higher risks of early childhood mortality, particularly during the neonatal period.

In other regions with a cultural preference for sons, periods of famine are associated with an increase in discrimination against female children reflected in; food distribution e.g. Bangladesh 1974-75 famine (Bairagi 1986; Bairagi and Langsten 1986); clinic attendance e.g. Northern Somalia 1984-85 (Watson, Wilkinson et al. 1985) and mortality patterns e.g. Ethiopia (Kidane 1990). These findings appear to contradict those predicted in the Trivers-Willard model; however, in Bangladesh the most marked anti-female biases occurred among the highest socio-economic status families, indicating that there may be intra-populational variation (Bairagi 1986). However, due to sample size constraints it is not possible to test for trends in sex biased mortality within sections of the sampled Oromo community.

126 Chapter 7: Sex biased investment

Table 7.7 Multivariate logistic regression modelling probability of dying in early childhood (<24 mths) by birth year cohort (1979-2000)

Variable Probability of death under 2 Odds ratio P Sex Male 0.560 0.023** Female - -

Year of birth 1979-80 1.886 0.004*** 1981-82 1.291 0.238 1983-84 1.175 0.446 1985-86 1.123 0.576 1987-88 1.093 0.674 1989-90 1.336 0.160 1991-92 -- 1993-94 0.612 0.032** 1995-96 0.539 0.007** 1997-98 0.340 0.000*** 1999-00 0.075 0.000***

Parity First birth 1.070 0.586

Later births --

Maternal age at birth <20 1.126 0.288 20-29 -- 30-39 0.828 0.029** 40-49 1.133 0.205

Sex* year of birth Sex*79-80 1.468 0.221 Sex*81-82 1.702 0.0951 Sex*83-84 1.998 0.025** Sex*85-86 1.969 0.025** Sex*87-88 1.456 0.226 Sex*89-90 1.202 0.549 Sex*91-92 -- Sex*93-94 1.606 0.152 Sex*95-96 2.108 0.021** Sex*97-98 1.192 0.063t Sex*99-00 7.255 0.000***

Intercept -1.506 0.000*** Births (n) 6382 Deaths 1037 Notes: fp=<0.1, **p=<0.05, ***p=<0.005

127 Chapter 7; Sex biased investment

Fig 7.9 Trends in infant mortality rates (1969-2000)

w 400 I 350 i 300 0 % 250 a> 200 I^ 150 1 100 c c

69-72 73-76 77-80 81-84 85-88 89-92 93-96 97-00

Year of birth

Fig 7.10 Proportion of childhood deaths (<24 months) for sex (1979-2000) [male = — , female = — - ]

0.35 0.3 0.25 0.2 drought drought 0.15

0.05

79 81 83 85 87 89 91 93 95 97 99 Year of birth

Fig 7.11 Relative sex ratio of neonatal and post-neonatal deaths (1979-200)

0.9

0.7 0.6 0.5 ■neonatal deaths 0.4 - - - post-neonatal deaths 79 81 83 85 87 89 91 93 95 97 99

128 Chapter 7: Sex biased investment

7.7 Concluding remarks In this chapter evidence of intra- and inter-populational variation in sex-ratio at birth (as predicted by the Trivers-Willard hypothesis) is presented. The results indicate that mothers in good condition bias their pre-natal investment towards sons, while those in poor conditions bias towards daughters. Producing male offspring under adverse ecological conditions is likely to be a risky strategy, since males are exposed to higher neonatal mortality risks, and for whom small body size confers a greater fitness cost than for females. Similar effects of maternal condition influencing sex-ratios may not have been found in other well-nourished human populations, where low birth weights and intra-uterine deaths are not generally related to maternal undemutrition, but to other aspects of socio-economic and health status, e.g. maternal smoking behaviour or proximity to environmental pollutants. However, under conditions of severe food shortage, as experienced periodically by the Oromo, a tendency to carry more females to term and to reduce the numbers of vulnerable low birth weight males would be adaptive.

There is no clear evidence for an adaptive model of differential sex-biased post-natal investment in this population. The overall state cultural preference is for sons. While male offspring are clearly more vulnerable than females during periods of acute food shortage, particularly during the neonate, there is no association between a current measure of resource availability (maternal condition) and mortality of the sexes. However, if intra-uterine selection has already occurred and the most robust males are selectively bom to the healthiest, well-provisioned mothers then sex-biased mortality is unlikely to be apparent.

7.7.1 Suggestions for future research A further topic for research, derived from the analyses presented in this and previous chapters, could be to investigate patterns of parental investment associated with the new development project. Increased family sizes in this population may result in increased competition between sibs for parental resources. This could have a detrimental influence not only in overall levels of investment (e.g. Madise and Diamond 1995; Bohler and Bergstrom 1995) but also on sex biases (Muhuri and Preston 1991). Girls may acquire new tasks, such as collecting water (which were previously considered too arduous) or increased childcare duties to closely spaced younger siblings. Increased energetic

129 Chapter 7; Sex biased investment stresses may cause a general increase in childhood malnutrition and morbidity, or the stated cultural preference for boys to become more evident through intra-household food allocation.

Collecting detailed time allocation data could provide a clearer picture of how women are using the time freed from water collection duties, including information on childcare practices; while anthropometric measurements and a detailed health survey from children could provide additional information on childhood growth patterns and health.

130 8

Conclusion

By combining theoretical models developed in evolutionary and reproductive ecology with demographic data collection techniques, this thesis has identified levels of population change in villages of rural Arsi. The analyses indicate that fertility may be increasing in response to a new development technology, despite other infrastructural and social developments, which are driving the secular trend towards increasing age at marriage. Contrary to expectations, the improved water supply and reduction in workloads have not immediately improved women’s health and body condition, or reduced the risks of early child death. However, a newly implemented programme of sanitation, and hygiene education may improve health and increase survivorship in the future.

Reduced birth intervals, in combination with a general downward trend in child mortality across the region, may increase family sizes. This is likely to place extra energetic stresses on both women and their offspring. A decline in maternal health and body condition, associated with repeated pregnancies and closely spaced births (maternal depletion syndrome), is evident in this population.

The Oromo of rural Arsi are a food-stressed population prone to periodic drought and crop failure, which result in food shortages. The nature of the nutritional inadequacies experienced by this population may explain some of the clear bio-demographic relationships demonstrated in this thesis, which have not been evident in other intra- populational studies, e.g. sex ratio adjustment (the Trivers-Willard effect) and parity- specific depletion in maternal condition. External factors, i.e. unique socio-ecological events, occurring in the region (e.g. sédentarisation and famine in the 1980s), may have prevented individuals from making behavioural and thus energy expenditure adjustments to low food intake. Additionally, the climate and geography of the villages Chapter 8: Conclusion has favoured lower levels of the diseases which have shaped demographic and health patterns in other regions of African (e.g. malaria in The Gambia, (Sear 2001), sexually- transmitted diseases in Botswana (Pennington and Harpending 1993)).

Labour-saving development projects aim to improve women’s welfare and alleviate the poverty of the entire community by reducing the time and energy that women expend on household chores. However, such intervention schemes may have unforeseen demographic consequences. By examining the dynamics involved in the timing of childbearing and mortality in rural Arsi this thesis has highlighted a number of broad bio-demographic issues which development policy-makers should be addressing when designing intervention schemes aimed at non-contracepting populations. An awareness that demographic changes could occur might influence both the behaviour of users and the nature of services offered in future intervention. Any increase in fertility associated with the water supply project, in a region characterised by declining agricultural productivity, is likely to contribute to an already great need for matemal-child health services and an increase in the currently low demand for contraception.

This thesis has identified the need for greater consideration of the demographic consequences of intervention schemes. Furthermore, that the direct and indirect effects of shorter birth spacing and increased family sizes may present serious health problems for women and children in transitional communities.

132 ______References

Barker, D. J. P. 1998. Mothers, Babies and Health in Later Life. Edinburgh: Churchill Livingstone. Barnes, D., L. Adair, and B. Popkin. 1991. Women's physical activity and pregnancy outcome: a longitudinal analysis from the Philippines. International Journal of Epidemiology 20:162-172. Barrett, H. R., and A. W. Browne. 1994. Women's time, labor-saving devices and rural- development in Africa. Community Development Journal 29:203-214. Baxter, P. 1996. "Towards a comparative ethnography," in Being and Becoming Oromo: Historical and Anthropological Enquiries. Edited by P. Baxter, J. Hultin, and A. Triulzi. Stockholm: Gotab. Beitens, I. Z., J. W. McArthur, B. A. Turnbull, 0. S. Skrinar, and B. A. Bullen. 1991. Exercise induces two types of hormone luteal dysfunction: confirmation by urinary free progesterone. Journal of Clinical Endocrinological Metabolism 72:1350-1358. Belay, N. "Arsi and Bale zones of Oromiya: Increasing vulnerability to food shortages," 1999. Addis Ababa: United Nations Development Programme - Emergencies Unit for Ethiopia (UNDP-EUE). http://www.telecom.net.et/~undp- eue/1999mnu.htm (March, 2002). Beneria, L. 1982. Women and Development: The Sexual Division o f Labour in Rural Societies. New York: Praeger. Bentley, G. R. 1985. Hunter-gatherer energetics and fertility: A reassessment of the !Kung San. Human Ecology 13:79-109. Bentley, G. R., V. J. Vitzthum, E. Caceres, H. Spielvogel, K. Crone, L. May and R. T. Chatterton. 1998. Reproduction and ecology in Provinicia Aroma, Bolivia. American Journal of Physical Anthropology 26 (supplement): 110. Bentley, G. R., R. Aunger, A. M. Harrigan, M. Jenike, R. C. Bailey, and P. T. Ellison. 1999. Women's strategies to alleviate nutritional stress in a rural African society. Social Science and Medicine 48:149-162. Berhane, Y. 2000. Women's health and reproductive outcome in rural Ethiopia. Medical dissertation. New Series No 674, Umea University, Sweden. Berhane, Y., Y. Gossaye, M. Emmelin, and U. Hogberg. 2001. Women's health in a rural setting in societal transition in Ethiopia. Social Science & Medicine 53:1525-1539.

133 References

Bock, J. 1999. Evolutionary approaches to population: Implications for research and policy. Population and Environment 21:193-222. Bohler, E., and S. Bergstrom. 1995. Subsequent pregnancy affects morbidity of previous child. Journal o f Biosocial Science 27:431-442. Bongaarts, J. 1993. "The relative contributions of biological and behavioural factors in determining natural fertility: a demographer's perspective," in Biomedical and Demographic Determinants o f Reproduction. Edited by R. Gray, H. Leridon and A. Spira. Oxford: Clarendon Press. Bongaarts, J., and M. Cain. Editors. 1982. Demographic Response to Famine. New York: GUP. Bongaarts, J., and R. 0. Potter. 1983. Fertility, Biology and Behaviour: An Analysis o f Proximate Determinants. Academic Press: London. Boone, J. 1988. "Parental investment, social subordination and population processes among the 15th and 16th century Portuguese nobility," in Human Reproductive Behaviour: A Darwinian Perspective. Edited by L. Betzig, M. Borgerhoff Mulder, and P. Turke, pp. 201-219. Cambridge: Cambridge University Press. Brabin, B. J., M. Hakimi, and D. Pelletier. 2001. An analysis of anemia and pregnancy- related maternal mortality. Journal o f Nutrition 131:604S-614S. Branca, F., 0. Pastore, T. Demissie, and A. Ferro-Luzzi. 1993. The nutritional impact of seasonality in children and adults of rural Ethiopia. European Journal of Clinical Nutrition 47:840-850. Bryceson, D. F., and J. Howe. 1993. Rural household transport in Africa: reducing the burden on women? World Development 21:1715-1728. Bullen, B., 0. S. Skrinar, I. Z. Beitens, G. Vonmering, B. A. Turnbull, and J. W. McArthur. 1985. Induction of menstrual disorders by strenuous exercise in untrained women. New England Journal o f Medicine 312:1349-1353. Bumpass, L. L., R. R. Rindfuss, and R. B. Janosik. 1978. Age and marital status at first birth and the pace of subsequent fertility. Demography 15:75-86. Buschkens, W., and L. Slikkerveer. 1982. Health Care in East Africa: Illness and Behaviour o f the Eastern Oromo in Hararghe (Ethiopia). Assen: Van Gorcum. Butz, W. P., J. P. Habicht, and J. Davanzo. 1984. Environmental-factors in the relationship between breastfeeding and infant-mortality - the role of sanitation and water in Malaysia. American Journal o f Epidemiology 119:516-525.

134 References

Caine, M. 1977. The economic activities of children in villages in Bangladesh. Population and Development Review 13:210-227. Campbell, K. L., and J. W. Wood. 1988. "Fertility in traditional societies," in Natural Human Fertility: social and biological mechanisms. Edited by P. Diggory, S. Teper, and M. Potts. London: Macmillan. Carr, M. 1985. "Technologies for rural women: impact and dissemination," in Technology for Rural Women. Edited by I. Ahmed, pp. 115-153. London: Allen and Unwin. Ceesay, S. M., A. M. Prentice, T. J. Cole, F. Ford, L. T. Weaver, E. M. E. Poskitt, and R. G. Whitehead. 1997. Effects on birth weight and perinatal mortality of maternal dietary supplements in rural Gambia: 5 year randomised controlled trial. British Medical Journal 315:786-790. Chacon-Puignau, G. C., and K. Jaffe. 1996. Sex ratio at birth deviations in modem Venezuela: The Trivers- Willard effect. Social Biology 43:257-270. Chen, L. C., S. Ahmed, M. Gesche, and W. H. Mosley. 1974. Prospective study of birth interval dynamics. Population Studies 28:277-297. Chisholm, M. 1968. Rural Settlement and Land Use: an essay in location, 2nd edition. Cambridge: CUP. Chowdhury, A. 1987. Changes in maternal nutritional status in a chronically malnourished population in Bangladesh. Ecology FoodNutr. 19:201-211. Clark, S., E. Colson, J. Lee, and T. Scudder. 1995. 10,000 Tonga - a longitudinal anthropological study from Southern Zambia, 1956-1991. Population Studies-a Journal of Demography 49:91-109. Clutton-Brock, T., and G. lason. 1986. Sex ratio variation in mammals. Quarterly Review of Biology 61:3 39-3 74. Cohen, J., and N.-I. Isaksson. 1987. Villagisation in the Arsi region o f Ethiopia. Upsala, Sweden: Swedish University of Agricultural Sciences. Report No. 19. Cooperstock, M., and J. Campbell. 1996. Excess males in preterm birth: Interactions with gestational age, race, and multiple birth. Obstetrics and Gynaecology 88:189-193. Copper, R. L., R. L. Goldenberg, R. K. Creasy, M. B. Dubard, R. O. Davis, S. S. Entman, J. D. lams, and S. P. Cliver. 1993. A multicenter study of preterm birth- weight and gestational-age specific neonatal-mortality. American Journal of Obstetrics and Gynaecology 168:78-84.

135 ______References

Cronk, L. 1989. Low socioeconomic-status and female-biased parental investment - the Mukogodo Example. American Anthropologist 91:414-429. —. 1991. Intention versus behavior in parental sex preferenees among the Mukogodo of Kenya. Journal o f Biosocial Science 23:229-240. CSA. 1998. "The 1994 Population and Housing Census of Ethiopia,", Vol. 1 edition. Addis Ababa, Ethiopia: Central Statistical Authority. CSA and ORC Macro. 2001. "Ethiopia Demographic and Health Survey 2000". Addis Ababa, Ethiopia and Calverton, Maryland, USA: Central Statistical Authority and ORC Macro. Cumming, D. 1993. "The effect of exercise and nutrition on the menstrual cycle," in Biomedical and Demographic Determinants of Reproduction. Edited by R. Gray, H. Leridon, and A. Spira. Oxford: Clarendon Press. Cumming, D., G. Wheeler, and V. Harber. 1994. "Physical activity, nutrition and reproduction," in Human Reproductive Ecology: Interactions of Environment, Fertility and Behaviour, vol. 709. Edited by K. Campbell and J. Wood, pp. 55- 76. New York: New York Academy of Sciences. Curtis, S., I. Diamond, and J. McDonald. 1993. Birth interval and family effects on postneonatal mortality in Brazil. Demography 30:33-43. Curtis, V. 1986. Women and the Transport o f Water. London: Intermediate Technology Press. Dagne, H. G. 1994. Early marriage in Northern Ethiopia. Reproductive Health Matters 4:35-38. Das Gupta, M. 1990. Death clustering, mother's education and the determinants of child mortality in rural Punjab, India. Population Studies 44:489-505. Davis, K., and J. Blake. 1956. Social structure and fertility: an analytic framework. Economic Development and Culture Change 4:211-235. De Sweemer, C. 1984. The influence of child-spacing on child survival. Population Studies 38:47-72. Delgado, H., R. Martorell, and R. Klein. 1982. Nutrition, lactation and birth interval components in rural Guatemala. American Journal of Clinical Nutrition 35:1468-1476. Desole, G., Y. Belay, and B. Zegeye. 1987. Vitamin-A-Deficiency in Southern Ethiopia. American Journal o f Clinical Nutrition 45:780-784.

136 ______References

Dickemann, M. 1979. "Female infanticide, reproductive strategies and social stratification," in Evolutionary Biology and Human Social Behaviour. Edited by N. Chagnon and W, Irons, pp. 321-369. Massachusetts: Duxbury Press. Dorjahn, V. 1959. Fertility, polygyny and their inter-relationships in Temne society. American Anthropologist 60:838-860. Dumin, J., and J. Womersley. 1974. Body fat assessed from total body density and its estimation from skinfold thickness: Measurements on 481 men and women aged from 16 to 72 years. British Journal o f Nutrition 32:77-97. Dyson, T. 1991. On the demography of South Asian famines .2. Population Studies-a Journal o f Demography 45:279-297. Dyson, T., and M. Murphy. 1985. The onset of fertility transition. Population and Development Review 11:399-440. Ellison, P. T. 1991. "Reproductive ecology and human fertility," in Applications of Biological Anthropology to Human Affairs. Edited by C. G. N. Mascie-Taylor and G. W. Lasker. Cambridge: Cambridge University Press. —. 1993. The ecological context of human ovarian function. Human Reproduction 8:2248-2258. — . 1994. Advances in human reproductive ecology. Annual Review of Anthropology 23:255-275. — . 2001. On Fertile Ground: A Natural History o f Reproduction. Cambridge, MA: Harvard University Press. Ellison, P. T., N. R. Peacock, and C. Lager. 1986. Salivary progesterone and luteal function in two low fertility populations in NE Zaire. Human Biology 58:473- 483. — . 1989. Ecology and ovarian function among Lese women of the Ituri forest. American Journal o f Physical Anthropology 78:519-526. Emlen, S., J. Emlen, and S. Levin. 1986. Sex-ratio selection in species with helpers-at- the-nest. American Naturalist 127:1-8. Esrey, S. A. 1996. Water, waste, and well-being: A multicountry study. American Journal of Epidemiology 143:608-623. Esrey, S. A., and J. P. Habicht. 1986. Epidemiologic evidence for health benefits from improved water and sanitation in developing-countries. Epidemiologic Reviews 8:117-128.

137 ______References

Esrey, S. A., J. B. Potash, L. Roberts, and C. Shiff. 1991. Effects of improved water- supply and sanitation on ascariasis, diarrhea, dracunculiasis, hookworm infection, schistosomiasis, and trachoma. Bulletin of the World Health Organisation 69:609-621. Eveleth, P., and J. Tanner. 1990. Worldwide Variation in Human Growth, 2nd edition. New York: Cambridge University Press. FAQ. "FAOSTAT Nutrition Data," 2001. Food and Agriculture Organisation. http://apps.fao/page/collections?subset=nutrition (March, 2002). Feachem, R. G., E. Bums, S. Caimcross, A. Cronin, P. Cross, D. Curtis et al. 1978. Water, Health and Development. London: Tri Med Books. Fedrick, J., and P. Adelstein. 1973. Influence of pregnancy spacing on outcome of pregnancy. British MedicalJournal 4:753-756. Feeney, G. A., and J. A. Ross. 1984. Analyzing open birth interval distributions. Population Studies-a Journal of Demography 38:473-478. Ferraz, E., R. Gray, P. Fleming, and T. Maia. 1988. Inter-pregnancy interval and low birth weight: findings from a case-control study. American Journal of Epidemiology 128:1111-1116. Ferro-Luzzi, A. 1990a. Seasonal energy deficiency in Ethiopian rural women. European Journal of Clinical Nutrition 44:7-18. —. 1990b. Seasonal energy stress in marginally nourished mral women. European Journal of Clinical Nutrition 44:41-46. Ferro-Luzzi, A., and W. P. T. James. 1996. Adult malnutrition: Simple assessment techniques for use in emergencies. British Journal of Nutrition 75:3-10. Ferro-Luzzi, A., S. Setter, M. Franklin, and W. P. T. James. 1992. A simplified approach of assessing chronic energy deficiency. European Journal o f Clinical Nutrition 46:173-186. Fisher, R. 1930. The Genetical Theory o f Natural Selection. Oxford: Clarendon Press. Forbes, R. 1925. From Red Sea to Blue Nile. New York: MacCaulay. Ford, K., S. Huffman, A. Chowdhury, S. Becker, and J. Menken. 1989. Birth interval dynamics in rural Bangladesh and maternal weights. Demography 26:425-437. Foster, A., J. Menken, A. Chowdhury, and J. Tmssell. 1986. Female reproductive development: a hazards model analysis. Social Biology 33:183-198.

138 ______References

Freedman, D., A. Thornton, D. Cambum, A. Duane, and L, Young-DeMarco. 1988. The life history calendar: a technique for collecting retrospective data. Sociological Methodology 88:37-68. Freij, L., and S. Wall. 1977. Exploring child health and its ecology. Acta Paediatrica Scandinavia (Suppl) 267:1-180. Fricke, T., and J. D. Teachman. 1993. Writing the names: marriage style, living arrangements and the first birth interval in Nepali society. Demography 30:175- 188. Frisancho, A. R. 1988. Nutritional anthropometry. Journal of American Dieticians Association 88:553-555. — . 1990. Anthropometric standards for the assessment of growth and nutritional status. Ann Arbor: University of Michigan Press. Frisch, R. E. 1982. Malnutrition and fertility. Science 215:1272-1273. Frisch, R. E., and J. W. MacArthur. 1974. Menstrual cycles: fatness as a determinant of minimum weight for height and adolescent events. Science 174:1148-1149. Gamachu, D. 1977. Aspects o f Climate and Water Budget in Ethiopia. Addis Ababa University. Gam, S. 1992. "The iron-deficiency anemias and their skeletal manifestations," in Diet, Demography and Disease: Changing Perspectives on Anemia. Edited by P. Stuart-Macadam and S. Kent. New York: Aldine de Gruyter. Gamer, P., T. Smith, M. Baea, D. Lai, and P. Heywood. 1994. Matemal nutritional depletion in a mral area of Papua-New-Guinea. Tropical and Geographical Medicine 46:169-171. Gaulin, S. J. C., and C. J. Robbins. 1991. Trivers-Willard effect in contemporary North- American society. American Journal of Physical Anthropology 85:61-69. Gilgen, D., C. G. N. Mascie-Taylor, and L. Rosetta. 2001. Intestinal helminth infections, anaemia and labour productivity of female tea pluckers in Bangladesh. Tropical Medicine & International Health 6:449-457. Golden, M. H. N. 1994. Is complete catch-up possible for stunted malnourished children. European Journal o f Clinical Nutrition 48:S58-S71. Goodkind, D., and L. West. 2001. The North Korean famine and its demographic impact. Population and Development Review 27:219.

139 References

Graham, M. J., U. Larsen, X. P. Xu, 1999. Secular trend in age at menarche in China: A case study of two rural counties in Anhui Province. Journal of Biosocial Science 31:2:257-267. Gubhaju, B. B., M. K. Choe, R. D. Retherford, and S. Thapa. 1987. Infant mortality trends and differentials in Nepal. Studies in Family Planning 18:23-31. Guerrant, R., M. de Souza, and M. Nations. 1996. "Health impacts of household water quality and availability," in At the Edge o f Development: Health Crises in a Transitional Society. Edited by R. Guerrant, M. de Souza, and M. Nations. Durham, NC: Carolina Academic Press. Gurri, F. D., G. B. Pereira, and E. F. Moran. 2001. Well-being changes in response to 30 years of regional integration in Maya populations from Yucatan, Mexico. American Journal of Human Biology 13:590-602. Haile, F. 1989. Women fuelwood carriers and the supply of household energy in Addis- Ababa. Canadian Journal of African Studies-Revue Canadienne Des Etudes Africaines 23:442-451. Hailemariam, A. 1991. Fertility levels and trends in Arsi and Shoa regions of Central Ethiopia. Journal o f Biosocial Sciences 23:387-400. Hailemariam, A., and H. Kloos. 1993. "Population," in The Ecology of Health and Disease in Ethiopia. Edited by H. Kloos and Z. Zein. Boulder, Colorado: Westview Press. Hatch, M., B. Levin, X. Shu, and M. Susser. 1998. Maternal leisure time exercise and timely delivery. American Journal o f Public Health 88:1528-1533. Hawkes, K., J. F. O'Connell, and Blurton Jones, N. G. 1997. Hadza women's time allocation, offspring provisioning and the evolution of long postmenopausal life spans. Current Anthropology 38:551-577. Hawkes, K., J. F. O'Connell, N. G. Blurton Jones, H. Alvarez, and E. L. Chamov. 1998. Grandmothering, menopause and the evolution of human life histories. Proceedings o f the Natural Academy o f Science USA 95: 1336-1339. Ministry of Health. 1999. "Health and health related indicators,". Addis Ababa, Ethiopia: Ministry of Health. Hebert, C., J. Bouyer, D. Collin, and I. Menger. 1986. Spontaneous abortion and inter­ pregnancy interval. European Journal of Obstetric and Gynaecological Reproductive Biology 22:125-132. Henry, L. 1961. Some data on natural fertility. Eugenics Quarterly 8:81-91.

140 ______References

Hill, K., and A. M. Hurtado. 1996. Ache Life History: the Ecology and Demography o f a Foraging People. New York: Aldine de Gruyter. Hobcraft, J. N., J. W. McDonald, and S. Rutstein. 1983. Child-spacing effects in infant and early child mortality. Population Index 49:585-618. Hobcraft, J. N., J. W. McDonald, and S. O. Rutstein. 1985. Demographic determinants of infant and early childhood mortality: a comparative analysis. Population Studies 39:363-385. Hobcraft, J., and M. Murphy. 1986. Demographic event history analysis: A selective review. Population Index 52:3-27. Howell, N. 1979. Demography o f the Dobe IKung. Academic Press: New York. Huffman, S. L., A. K. M. A. Chowdhury, J. Chakraborty, and N. K. Simpson. 1980. Breastfeeding patterns in rural Bangladesh. American Journal of Clinical Nutrition 33:144-153. Huffman, S. L., H. A. Chowdury, and L. Nahar. 1987. Nutrition and fertility in Bangladesh: breast-feeding and post-partum amenorrhoea. Population Studies 41:447-462. Huffman, S. L., E. R. Zehner, and C. Victora. 2001. Can improvements in breast­ feeding practices reduce neonatal mortality in developing countries? Midwifery 17:80-92. Hughes, J. 1983. Potential impacts of improved water supply and excreta disposal on diarrhoeal disease morbidity: An assessment on a review of published studies. Geneva: WHO Diarrhoeal Disease Programme. Hurtado, A. M., and K. R. Hill. 1990. Seasonality in a forging society: variation in diet, work effort, fertility and sexual division of labour among the Hiwi of Venezuela. Journal o f Anthropological Research 46:293-346. Huttly, S., C. Victora, F. Barros, and J. Vaughan. 1992. Birth spacing and child health in urban Brazilian children. Pediatrics 89:1049-54. Jakobovits, A., A. A. Jakobovits, and A. Viski. 1987. Sex-Ratio of the stillborn fetuses and neonates dying in the 1st week. Early Human Development 15:131-135. James, W. P. T., A. Ferro-Luzzi, and J. Waterlow. 1988. Definition of chronic energy deficiency in adults. European Journal o f Clinical Nutrition 42:969-981. Jasienska, G. 2001. "Why energy expenditure causes reproductive suppression in women: An evolutionary and bio-energetic perspective," in Reproductive

141 References

Ecology and Human Evolution. Edited by P. T. Ellison. New York: Aldine de Gruyter. Jasienska, G., and P. T. Ellison. 1993. Heavy workload impairs ovarian function in Polish peasant women. American Journal of Physical Anthropology 16:117-118. —. 1998. Physical work causes suppression of ovarian function in women. Proceedings of Royal Society of London B 265:1847-1851. John, A. M., J. A. Menken, and A. K. M. A. Chowdhury. 1987. The effects of breast­ feeding and nutrition on fecundability in rural Bangladesh: a hazards-model analysis. Population Studies 41:433-446. Johnson, C. 1988. Dispersal and the sex ratio at birth in primates. Nature 332:726-728. Jones, C. D. R. 1989. Energy-Cost of Carrying Loads. European Journal of Clinical Nutrition 43:881-883. Jones, R. E., and A. Palloni. 1989. Breast-feeding and postpartum amenorrhoea in Indonesia. Journal o f Biosocial Science 21:83-100. Kebede, H. 1978. Improving Village Water Supplies in Ethiopia: A case-study o f socio­ economic implications. ECA/ UNICEF. Keller, M., R. Nesse, and S. Hofferth. 2001. The Trivers-Willard hypothesis of parental investment: No effect in the contemporary United States. Evolution and Human Behavior 22:343-360. Kellokumpulehtinen, P., and L. J. Pelliniemi. 1984. Sex-ratio of human conceptuses. Obstetrics and Gynecology 64:220-222. Khan, K. S., P. F. W. Chien, and N. B. Khan. 1998. Nutritional stress of reproduction - A cohort study over two consecutive pregnancies. Acta Obstetricia Et Gynecologica Scandinavica 77:395-401. Kidane, A. 1989. Demographic consequences of the 1984-1985 Ethiopian famine. Demography 26:515-522. — . 1990. Mortality estimates of the 1984-85 Ethiopian famine. Scandinavian Journal of Social Medicine 18:281-286. Kirchengast, S., and B. Hartmann. 1998. Maternal prepregnancy weight status and pregnancy weight gain as major determinants for new-born weight and size. Annals of Human Biology 25:17-28. Kline, J., Z. Stein, and M. Susser. 1989. Conception to Birth: Epidemiological Study o f Prenatal Development. Oxford: Oxford University Press.

142 References

Kloos, H., G. Desole, and A, Lemma. 1981. Intestinal parasitism in seminomadic pastoralists and subsistence farmers in and around irrigation schemes in the Awash Valley, Ethiopia, with special emphasis on ecological and cultural associations. Social Science & Medicine Part B-Medical Anthropology 15:457- 469. Kloos, H., and B. Lindtjom. 1994. Malnutrition and mortality during recent famines in Ethiopia - implications for food aid and rehabilitation. Disasters 18:130-139. Koenig, M. A., J. F. Phillips, O. Campbell, and S. D'Souza. 1990. Birth intervals and childhood mortality in rural Bangladesh. Demography 27:251-265. Koziel, S., and S. Ulijaszek. 2001. Waiting for Trivers and Willard: Do the rich really favor sons? American Journal of Physical Anthropology 115:71-79. Kramer, K, L., and G. P. McMillan. 1998. How Mayan females respond to changing technology. Human Nature 9:205-223. — . 1999. Women's labour, fertility and the introduction of modem technology in a rural Maya village. Journal of Anthropological Research 55:499-520. Krebs, J. R., and Davies, N. B. 1978. Behavioural Ecology - an evolutionary approach. Oxford: Blackwell Scientific Press. Kroeger, A. 1983. Health interview surveys in developing countries: A review of the methods and results. International Journal of Epidemiology 12:465-481. Kunstadter, P. 2001. "Emerging health threats associated with development: Hmong highlanders in Thailand." Conference Presentation, International Union for the Scientific Study of Population (lUSSP^, Salvador, 2001. Kunstadter, P., S. L. Kunstadter, P. Leepreecha, C. Podhisita, L. Y. Mai, C. S. Thao, R. S. Thao, and W. S. Yang. 1992. Causes and consequences of increase in child survival rates - Ethnoepidemiology among the Hmong of Thailand. Human Biology 64:821-841. Kwast, B., and J. Stevens. 1987. Viral hepatitis as a major cause of maternal mortality in Addis Ababa. International Journal o f Gynaecology & Obstetrics 25:99-106. Lager, C., and T. Ellison. 1990. Effect of moderate weight loss on ovarian function assessed by salivary progesterone measurements. American Journal of Human Biology 2:303-312. Le Grand T. K. and J. F. Phillips. 1996. The effect of fertility reductions on infant and child mortality: Evidence from Matlab in rural Bangladesh. Population Studies- a Journal of Demography 50:51.

143 ______References

Leslie, P. W., K. L. Campbell, and M. A. Little. 1993. Pregnancy loss in nomadic and settled women in Turkana, Kenya: a prospective study. Human Biology 65:237- 254. Leslie, P. W., and P. H. Fry. 1989. Extreme seasonality of births among nomadic Turkana pastoralists. American Journal of Physical Anthropology 79:103-115. Levine, D. 1965. Wax and Gold: Tradition and Innovation in Ethiopian Culture. Chicago: University of Chicago Press. Lindstrom, D. P., and B. Berhanu. 2000. The effects of breastfeeding and birth spacing on infant and early childhood mortality in Ethiopia. Social Biology 47:1-17. Little, M., and P. Leslie. 1999. Turkana Herders o f the Dry Savanna: Ecology and biobehavioural response o f nomads to an uncertain environment. Oxford: GUP. Little, M. A., P. W. Leslie, and K. L. Campbell. 1992. Energy reserves and parity of nomadic and settled Turkana women. American Journal of Human Biology 4:729-738. Lohman, T. G., A. F. Roche, and R. Martorell. 1988. Anthropometric Standardization Reference Manual. Champaign, Illinois: Human Kinetic Books. Low, B. S. 1993. Ecological demography: a synthetic focus in evolutionary anthropology. Evolutionary Anthropology 1:177-187. Low, B. S., C. Simon, and K. Anderson. 2002. An evolutionary ecological perspective on demographic transitions: modeling multiple currencies. American Journal of Human Biology 14:149-167. Lukmanji, Z. 1992. Women's workload and their impact on their health and nutritional status. Progress in Food and Nutrition Science, 16:12:163-179. Lucas, A., R. Morley, and T. J. Cole. 1998. Randomised trial of early diet in preterm babies and later intelligence quotient. British Medical Journal 317:1481-1487. Lucas, D., and J. Kane. 1985. Asking Demographic Questions. Demographic Teaching Notes. Canberra: National Centre for Development Studies, Australian National University. Maconochie, N., and E. Roman. 1997. Sex ratios: are there natural variations within the human population? British Journal of Obstetrics and Gynaecology 104:1050- 1053. Madise, N. J., and I. Diamond. 1995. Determinants of infant-mortality in Malawi - an analysis to control for death clustering within families. Journal of Biosocial Science 27:95-106.

144 References

Maloiy, G., N. Heglimd, L. Prager, G. Cavagna, and C. Taylor. 1986. Energetic cost of carrying loads: have African women discovered an economic way? Nature 319:668-669. Mammo, A., and S. P. Morgan. 1986. Childlessness in rural Ethiopia. Population and Development Review 12:533-546. Mascie-Taylor, C. G. N. 1996. "The relationship between disease and subfecundity," in Variability in Human Fertility. Edited by L. Rosetta and C. G. N. Mascie- Taylor. Cambridge: Cambridge University Press. Matte, T. D., M. Bresnahan, M. D. Begg, and E. Susser. 2001. Influence of variation in birth weight within normal range and within sibships on IQ at age 7 years: cohort study. British Medical Journal 323:310-314. Maynard Smith, J. 1980. A new theory of sexual investment. Behavioral Ecology and Sociobiology 7:247-251. McFalls, J. A., Jr., and M. H. McFalls. 1984. Disease and Fertility. New York: Academic Press. Merchant, K., R. Martorell, and J. D. Haas. 1990. Consequences for maternal nutrition of reproductive stress across consecutive pregnancies. American Journal of Clinical Nutrition 52:616-620. Miller, J. E. 1989. Determinants of intrauterine growth-retardation - Evidence against maternal depletion. Journal of Biosocial Science 21:235-243. Miller, J. E., and R. Huss-Ashmore. 1989. Do reproductive patterns affect maternal nutritional status? An analysis of maternal depletion in Lesotho. American Journal o f Human Biology 1:409-419. Miller, J. E., G. Rodriguez, and A. R. Pebley. 1994. Lactation, seasonality, and mothers postpartum weight change in Bangladesh - an analysis of maternal depletion. American Journal o f Human Biology 6:511-524. Moore, S. E., T. J. Cole, A. C. Collinson, E. M. E. Poskitt, 1. A. McGregor, and A. M. Prentice. 1999. Prenatal or early postnatal events predict infectious deaths in young adulthood in rural Africa. International Journal of Epidemiology 28:1088-1095. Moore, S. E., T. J. Cole, E. M. E. Poskitt, B. J. Sonko, R. G. Whitehead, 1. A. McGregor, and A. M. Prentice. 1997. Season of birth predicts mortality in rural Gambia. Nature 388:434-434.

145 References

Morris, S., C. Victoria, F. Barros, R. Halpem, A. Menezes, J. Cesar, B. Horta, and E. Tomasi. 1998. Length and pondéral index at birth: Associations with mortality, hospitalisations, development and post-natal growth in Brazilian infants. International Journal o f Epidemiology 27:242-247. Muhuri, P. K., and S. H. Preston. 1991. Effects of family composition on mortality differentials by sex among children in Matlab, Bangladesh. Population and Development Review 17:415-434. Nager, R. G., P. Monaghan, R. Griffiths, D. C. Houston, and R. Dawson. 1999. Experimental demonstration that offspring sex ratio varies with maternal condition. Proceedings of the National Academy of Sciences of the United States o f America 96:570-573. Nath, D. C., K. C. Land, and G. Goswami. 1999. Effects of the status of women on the first birth interval in Indian urban society. Journal of Biosocial Science 31:55- 69. Nath, D. C., K. K. Singh, K. C. Land, and P. K. Talukdar. 1993a. Age of marriage and length of the first birth interval in a traditional Indian society: life table and hazards model analysis. Human Biology 65:783-797. Nath, D. C., K. K. Singh, K. C. Land, and P. K. Talukdar. 1993b. Breastfeeding and Postpartum Amenorrhoea in a Traditional Society: A Hazards Model Analysis. Social Biology 40: 74-86. Newman, J. 1995. The Peopling o f Africa, Ch. 6. Newhaven, CT: Yale University Press. Nonaka, K., B. Desjardins, H. Charbonneau, L. Legare, and T. Miura. 1998. Marriage season, promptness of successful pregnancy and first-born sex ratio in historical natural fertility population - evidence for sex-dependent early pregnancy loss? International Journal o f Biometeorology 42:89-92. Palloni, A., and M. Tienda. 1986. The effects of breast-feeding and pace of childbearing on mortality at early ages. Demography 23:31-52. Panter-Brick, C. 1989. Motherhood and subsistence work: The Tamang of rural Nepal. Human Ecology 17:205-228. — . 1991. Lactation, birth spacing and maternal work-loads among two casts in rural Nepal. Journal o f Biosocial Sciences 23:137-154. Panter-Brick, C., D. S. Lotstein, and P. T. Ellison. 1993. Seasonality of reproduction function and weight loss in rural Nepali women. Human Reproduction 8:684- 690.

146 References

Parker, B. 1995. Ethiopia: Breaking New Ground. Oxfam. Pasquini, P., L. Bisanti, L. Soldo, P. Palladino, C. Rozera, E. Frontini, A. Miozzo, T. M. Digennaro, A. Indi, S. Conti, and M. Rapicetta. 1988. Hepatitis-B infections in the Arsi Region of Ethiopia. European Journal of Epidemiology 4:310-313. Pastore, G., F. Branca, T. Demissie, and A. Ferro-Luzzi. 1993. Seasonal energy stress in an Ethiopian rural community: an analysis of the impact at the household level. European Journal o f Clinical Nutrition 47:851-862. Pawlowski, B., R. Dunbar, and A. Lipowicz. 2000. Evolutionary fitness: Tall men have more reproductive success. Nature 403:156. Peacock, N. 1991. An evolutionary perspective on the patterning of maternal investment in pregnancy. Human Nature 2:351-385. Pebley, A. R., A. I. Hermalin, and J. Knodel. 1991. Birth spacing and infant mortality: Evidence for 18th and 19th century German villages. Journal of Biosocial Science 23:445-59. Pedersen, J. 2000. Determinants of infant and child mortality in the West Bank and Gaza Strip. Journal o f Biosocial Science 32:527-546. Pennington, R., and H. Harpending. 1993. The Structure of an African Pastoralist Community: Demography, History and Ecology o f the Ngamiland Herero. Oxford: Clarendon Press. Phillips, D. I. W., D. J. Handelsman, J. G. Eriksson, T. Forsen, C. Osmond, and D. J. P. Barker. 2001. Prenatal growth and subsequent marital status: longitudinal study. British Medical Journal 322:771-771. Prentice, A. M., R. G. Whitehead, S. B. Roberts, and A. A. Paul. 1981. Long-term energy-balance in childbearing Gambian Women. American Journal of Clinical Nutrition 34:2790-2799. Pressât, R. 1985. The Dictionary o f Demography. Oxford: Blackwell. Raikest, A. 1989. Women's health in East Africa. Social Science & Medicine 28:447- 459. Ramakrishnan, U., H. Barnhart, D. G. Schroeder, A. D. Stein, and R. Martorell. 1999. Early childhood nutrition, education and fertility milestones in Guatemala. Journal o f Nutrition 129: 12: 2196-2202. Rashid, M. J., and S. J. Ulijaszek. 1999. Daily energy expenditure across the course of lactation among urban Bangladeshi women. American Journal of Physical Anthropology 110:457-465.

147 ______References

Rawlings, J. S., V. B. Rawlings, and J. A. Read. 1995. Prevalence of low-birth-weight and preterm delivery in relation to the interval between pregnancies among White and Black-women. New England Journal o f Medicine 332:69-74. Read, J. S., J. F. Troendle, and M. A. Klebanoff. 1997. Infectious disease mortality among infants in the United States, 1983 through 1987. American Journal of Public Health 87:192-198. Retherford, R. D., M. K. Choe, S. Thapa, and B. B. Gubhaju. 1989. To what extent does breastfeeding explain birth interval effects on early childhood mortality. Demography 26:439-450. Rivers, J., and M. Crawford. 1974. Maternal nutrition and the sex ratio at birth. Nature 252:297-8. Roff, D. 1992. The Evolution o f Life Histories: Theory and Analysis. New York: Chapman Hall. Rosetta, L. 1993. Female reproductive dysfunction and intense physical training. Oxford Review o f Reproductive Biology 115:113-141. Ross, D., and J. Vaughan. 1986. Health interview surveys in developing countries: A methodological review. Studies in Family Planning 17:78-94. Sear, R. 2001. Evolutionary demography of a rural Gambian population. University College London. Unpublished PhD Thesis. Sear, R., R. Mace, and I. A. McGregor. 2000. Maternal grandmothers improve nutritional status and survival of children in rural Gambia. Proceedings o f the Royal Society of London Series B-Biological Sciences 267:1641-1647. Sear, R., F. Steele, A. A. McGregor, and R. Mace. 2002. The effects of kin on child mortality in rural Gambia. Demography 39:43-63. Selinus, R., A. Gobezie, K. E. Knutsson, and B. Vahlquist. 1971. Dietary studies in Ethiopia: dietary pattern among the Rift Valley Arsi Galla. American Journal of Clinical Nutrition 24:365-377. Shamebo, D., A. Sandstrom, L. Muhe, L. Freij, I. Krantz, G. Lonnberg, and S. Wall. 1993. The Butajira Project in Ethiopia - a nested case-referent study of under-5 mortality and its public health determinants. Bulletin o f the World Health Organization 71:389-396. Shangold, M., R. Freeman, B. Thysen, and M. Gatz. 1979. The relationship between long-distance running, plasma progesterone and luteal phase length. Fertility and 31:130-133.

148 References

Silkin, T. 1998. Hitosa Water Supply: A people's project. London: Water Aid. Slikkerveer, L. 1990. Plural Medical Systems in the Horn of Africa. London: Paul Kegan International. Srinivasan, K. 1972. Analytical Models for Two Types o f Birth Intervals with Applications to an Indian Population. Tivandrum, India: St Joseph's Press. Steams, S. C. 1992. The Evolution o f Life Histories. Oxford: Oxford University Press. Stein, Z., M. Susser, G. Saenger, and F. Marolla. 1975. Famine and Human Development: The Dutch Hungry Winter o f 1944-45. New York: Oxford University Press. Strassman, B., and B. Gillespie. 2002. Life-history theory, fertility and reproductive success. Proceedings of the Royal Society of London Series B-Biological Sciences 269:553-562. Strassman, B. I., and J. H. Warner. 1998. Predictors of fecundability and conception waits among the Dogon of Mali. American Journal of Physical Anthropology 105:167-184. Tafari, N., R. L. Naeye, and A. Gobezie. 1980. Effects of maternal undemutrition and heavy physical work during pregnancy on birth weight. British Journal of Obstetrics and Gynaecology 87:222-226. Taffesse, S., and W. Girma. "Maternal energy and macronutrient insecurity in an enset- com staple village of [Sidama] Southern Ethiopia," 1996-2000: Sidama Concern, http://sidamaconcem.com (December, 2001). Teka, G. 1993. "Water supply and sanitation," in The Ecology of Health and Disease in Ethiopia. Edited by H. Kloos and Z. Zein, Ch. 14. Boulder, Colorado: Westview Press. Terefe, H. 2000. A Study of Female Genital Mutilation and Reproductive Health: The Case of the Arsi Oromo, Ethiopia. University of Gottingen. Unpublished PhD thesis. Thomson, A. 1959. Diet in pregnancy. British Journal of Nutrition 13:509-525. Tracer, D. P. 1991. Fertility-related changes in matemal body-composition among the Au of Papua-New-Guinea. American Journal of Physical Anthropology 85:393- 405. —. 1994. Socioeconomic mediators of birth interval duration in a Papua New Guinea population. Annals o f the New York Academy o f Science 709:231-233.

149 ______References

—. 1996. Lactation, nutrition, and postpartum amenorrhoea in lowland Papua New Guinea. Human Biology 68:277-292. Trimingham, J. 1965. Islam in Ethiopia. London: Frank Cass & Co. Ltd. Trivers, R. 1972. "Parental investment and sexual selection," in Sexual Selection and the Descent o f Man. Edited by B. Campbell, pp. 136-179. London: Heinemann. Trivers, R., and D. Willard. 1973. Natural selection of parental ability to vary the sex ratio of offspring. Science 179:90-92. Trussell, J., and J. Menken. 1978. Early child-bearing and subsequent fertility. Family Planning Perspectives 10:209-218. Tuma, N., and M. Hannan. 1978. "Approaches to censoring problem in analysis of event histories," in Sociological Methodology. Edited by K. Schuessler. San Francisco: Jossey-Bass. Ulijaszek, S. J. 1993. "Seasonality in reproductive performance in rural Gambia," in Seasonality and Human Ecology. Edited by S. J. Ulijaszek and S. S. Strickland, Ch. 7. Cambridge: Cambridge University Press. Ulijaszek, S. J., and S. S. Strickland. 1993. "Nutritional studies in biological anthropology," in Research Strategies in Human Biology: Field and Survey Studies. Edited by G. W. Lasker and C. G. N. Mascie-Taylor, Ch. 6. Cambridge: Cambridge University Press. UNICEF. 1993. Children and Women in Ethiopia: A Situation Report. Vitzthum, V. J. 1989a. Breast-feeding patterns in rural highland Peruvian community. American Journal of Physical Anthropology 69:275. —. 1989b. Nursing behaviour and its relation to duration of post-partum amenorrhoea in an Andean community. Journal o f Biosocial Sciences 21:145-160. — . 1997. "Flexibility and paradox: The nature of adaptation in human reproduction," in The Evolving Female: A Life-History Perspective. Edited by M. Morbeck, A. Galloway, and A. Zihlman, pp. 242. Princeton, New Jersey: Princeton University Press, vom Saal, F. S., and C. E. Finch. 1994. "Reproductive senescence," in The Physiology o f Reproduction, 2nd edition. Edited by E. N. Knobil, JD, pp. 1213-1314. New York: Raven Press. Watson, R., R. Wilkinson, and N. Greenberg. 1985. We are the children. Newsweek:2S- 34.

150 ______References

Wenlock, R. J., and R. W. Wenlock. 1981. Matemal nutrition, prolonged lactation and birth-spacing in Ethiopia. Journal o f Biosocial Sciences 13:261-268. White, G., D. Bradley, and A. White. 1972. Drawers o f Water: Domestic Water Use in East Africa. Chicago: Chicago University Press. Winikoff, B., and M. Castle. 1988. The matemal depletion syndrome: clinical diagnosis of eco-demographic condition. Biology and Society 5:163-170. Winkler, E. M., and S. Kirchengast. 1994. Body dimensions and differential fertility in Kung San males from Namibia. American Journal o f Human Biology 6:203-213. Winkvist, A., K. M. Rasmussen, and J. P. Habicht. 1992. A new definition of matemal depletion syndrome. American Journal o f Public Health 82:691-694. Winkvist, A., F. Jalil, J. P. Habicht, and K. M. Rasmussen. 1994. Matemal energy depletion is buffered among malnourished women in Punjab, Pakistan. Journal o f Nutrition 124:2376-2385. Winkvist, A., J. P. Habicht, and K. M. Rasmussen. 1998. Linking matemal and infant benefits of a nutritional supplement during pregnancy and lactation. American Journal o f Clinical Nutrition 68:656-661. Wood, J. W. 1990. Fertility in anthropological populations. Annual Review of Anthropology 19:211-242. —. 1994. Dynamics o f Human Reproduction: Biology, Biometry and Demography. New York: Aldine de Gmyter Inc. Wood, J. W., D. Lai, P. L. Johnson, K. L. Campbell, and I. A. Maslar. 1985. Lactation and birth-spacing in highland Papua New Guinea. Journal of Biosocial Sciences 9(Suppl.):159-173. Yamaguchi, K. 1991. Event History Analysis. Thousand Oaks, Califomia: Sage Publications. Yasmin, S., D. Osrin, E. Paul, and A. Costello. 2001. Neonatal mortality of low-birth- weight infants in Bangladesh. Bulletin o f the World Health Organization 79:608-614. Yohannes, A. G., K. Streatfield, and L. Bost. 1992. Child morbidity pattems in Ethiopia. Journal of Biosocial Science 24:143-155.

151 Appendix 1: Illustrations

1. Arsi rift valley landscape

2. Peasant association leader and militia, Reissa Michiko Appendix 1: Illustrations

Local ponds dry up in the dry season

f

4. Collecting pond water Appendix 1: Illustrations

i t

5. Transporting w ater In in sera 6. Filling in sera at the tap

7. Queuing for water at the tap stand Appendix 1: Illustrations

K -, X ’ “ IT', ' - i k

8. Field staff November 1999 - June 2000

9. Collecting a birth history 10. Measuring height Appendix 1: Illustrations

11. Unmarried girls performing at a Muslim wedding

12. Well-nourished child in Terro Moyee Appendix 1: Illustrations

13. Arsi Oromo (Muslim) woman dressed for market

14. Shoa Oromo (Orthodox Christian) woman and family Appendix II: Regional dates

Regional Historical Dates (adapted from Parker, 1995)

1930 Haile Selassie crowned emperor of Ethiopia; rural areas under traditional feudal system

1936 Italy invades Ethiopia, followed by six years of military occupation

1941 Italians defeated by a guerilla army and Allied troops

1962 Eritrea annexed by Haile Selassie, prompting emergence of an Eritrean rebel movement (ELF); the start of the 30 year civil war

1972-74 Famine kills 200,000

1974 The Derg’ overthrows an aging and unpopular Haile Selassie

1977 Lt. Col. Mengistu Hailemariam emerges as the undisputed leader the the Derg regime

1980s Throughout this decade the Derg implement a number of land reform intiatives: A. resettlement programmes: 600,000 rural inhabitants are moved, premoted as a long-term solution to population pressure and environmental degradation in the highlands B. villagisation programmes: 3rd of the rural population are relocated from scatterd homesteads into modern villages 0 . land redistribution programmes: private ownership of land is abolished making land the collective property of the people; individual farmers are allocated up to 10 hectares each, producer co-operatives formed

1984-85 An estimated 0.5 million die in nation-wide famine

1986 Mengistu establishes the People’s Democratic Republic of Ethiopia, with a Marxist-Leninist constitution

1989 Government is driven out of the capital of Tig ray province by rebel movement TPLF (Tigrayan People’s Liberation Front); rebel coalition ERDF (Ethiopian People’s Revolutionary Democratic Front) is formed from TPLF and other rebel groups from other regions, including Eritrea

May 1991 Mengistu flees country before EPRDF enter Addis Ababa and EPLF enters Asmara; transitional government re-establishes order

1992 WaterAid undertakes a baseline survey of Hitosa woreda, Arsi province

April 1993 98% of Eritreans vote for independence, which is granted; borders not clearly defined

Oct 1993 WaterAid and Water Action commence Phase I of the Hitosa Gravity Water Supply Scheme; engineering work starts

July 1996 Phase I of Hitosa Gravity Water Supply Scheme completed

1999- Phase II of the water supply scheme commences and plans made to extend the pipeline and supply scheme into adjacent area between Gonde and Iteya.

May-Jun 1998 Hundreds die in border fighting between Ethiopia and Eritrea; conflict halted as rainy season begins Appendix II: Regional dates

January 1999 Isolated artillery fire in border region

February 1999 UN peace mission to both countries; OAU peace plan proposed; heavy ighting resumes on three fronts (Badame, Bure and Zalambesa); casualties on both sides estimated to be in the 100,000’s; world’s largest conventional war

June 1999 Beig (short rains) fail; south-east and southern Ethiopia experience famine. ID Dumbcr ? VILLAGE: HEAD OF HOUSEHOLD: . 1. 1 .1 J .1 I a DATE: NAME OF RESPONDENT: . "4,o TIME: ETHNIC GROUP/ REUCION:. 3" O Tbcse quctlioM refer to the whole hooeehoM le. cvcryoM who eoreielly live» le the houehold. 5 DEMOGRAPHIC INFORMATION FOR ALL PERSONS TO BE ASKED OF THE HOUSEHOLD HEAD: FOR EVER MARRIED WOMEN: N Sea Relationship - Age- Father Education - Residence Movement - Age at Wife Rank - Children at Children Children Check 6 patterns - elsewhere- 0 Name of everyone Mor Identify closest How old No. of years of Wbal was Is What rank wife So. currently living in your F household is...? Is their Is their school have they How long has his/her main currently is...... ? How many How many How many total 5 household. relationship birth birth completed? ...lived in this married(M), How many co­ children tlut children that you of children (yeon) father village moving to this widowedfW) fCrve wives does you gave birth gave birth to are of (Slarl wUh head of (eg. son of 2*3 alive? alive? (include adult continuously? village? separatedfS), age In .... ? to are currently currently living gave birth to household) A •Adopted educ.. primary, (Years and divorced(D), years) living at home? elsewhere? have died? son if.....) Y/N Y/N secondary) (Record as ....? State Koraele A— always mairiedfN)? Rank/Toial) M F M F M F 1 1. 2. I 3. §■ 4. ?

6. - 7. a 9. I to

II

12

13 14 0 S HOUSEHOLD STATUS 3. Does any member ofyour household own any of the following? CO I Whit is Ifae main ecoooniic activity (in tenu ofthne (pent in Ihel activity) of the boinehold? OIL LAMP YES NO WOODEN BED YES NO table YES NO RADIO YES NO PLOUGH YES NO BICYCLE YES NO 2. Does anyone in your household own any llveatock? YES DONKEY CART YES NO CAR YES NO 1 If YES, what and how many 4. What kind of toilet fociihy does your household have? (Circk one) FLUSH TOILET / TRAD. PIT LATRINEfTPL) / VENTILATED IMPROVED PIT(VIP) / NO FACILITY I Please thank the respondentfor their time. Village, HHH I D n u m b e r : Date N a m e o f re s p o n d e n t _ I I I I I I Time Gosa keessan A m a n t i i r Q. Gaafilecn nmina isnii gafadhuii wnn’ee matii keessani ta’a - namoot mana kana kecssa jirataan O hiindaaf. Ragaalce iliiitmouirnphii kuiii namoota mana keessa jiraan bunda waa’ec malt ittiin 3" gaafalamiiiidha: Abtiaa warraa / Hadhaa warraa 0

.Sea Relationship- At*- Education - Residence patterns Movement - Age at First Wife Rank - Children at Children Children Check __ , bangs kutaa ___ . Sabnbonni marriage __ .abba* elsewhere - Eega, Maqaa namoota am ma Fiinimma warn Umrifal meeqaatii __ , osoo addaan guddaan manastieefnitti Ijooleen ____ Ijooleen __ baayinni niana kec keessa M or mana addaan Haaü_ Abbaan him cbtiin . naannoo ganda mceqahdha? deessee keessa deessee kecssa ijooleen /kee / 10 jiialaon. F kee Geessoda wagaa/ban meeqaa kanatli galcef Fiidhc/lee/ Abbaan kan amma kan amma id o o keessa kan _deesae kee fakk. Ilmoan meeqa B.sad. Iffaa ganda kana jiretee. /teef malii? jiraanii / heetu me/tee warashee/atti mana kana biiTOO liaan 5 Ahhoa mono iirti 2+3.. /< - ata’a? jhtil? hibuun mootumma/kan (yrs/mths.) Yrs/mths iimrisashee wajjin nitti keessa jiraalani jechuudh ialt/ahil GyJu'ifaeha blears) Y/N jini? amantii. (7.) A’-yeroo htmda A - Alwo)is UiH'/S/ meeqa lure? meeqaa qaba/da? a? U) (Î-) (3.). (4) (Ï-) Y/N («.) State Koranic as.) (9.) O/AIIO.) (II.) Sad/ldaXil.) M(t3.) F(td.) M(15.) F(l«.) M(t7.)F(ia.) (19.) 1

4. 5. I 6.

7.

8 ?

9. I I § II I 12

11

HOUSEHOLD STATUS 22. Qabeenya aai gadii kanaa keessa mana qilxhiu? 20. Abba mana yeroo hctkluu dalagaa miali irraltl dabmaa? Fbv Qollsa, banllam — KKF. Faanunsii / Kiirrazii YES NO Siree mukaa YES NO Xarapheezaa YES NO Raadiyoo YES NO I Maanishaa YES NO Bishkililtii YES NO 21. Mana kana kccsaa watrii kan horri/loon qabaanii jinni? YES NO Canii haice YES NO Konkoolata / Makinna YES NO ViMiJimix’. Kan eCTviiutll? Akaakuu boni? Mreaaa? 23. Mana Ticaanni kan ai.kamin: qabdnti? (Circleone) Ï F!,USH TOÎI.ET / TRAD. PIT LATRINE(TPL) / VENTIIATED IMPROVED PIT(VIP) / NO FACILITY Q) Galaloomal §• CO ID number EMW EVENT HISTORY CALENDAR; Interviewer Date CE The next set of questions are for EMW who have EMW; HH; Village: had a birth during the LAST 7 YEARS ONLY. Thank all other women for their time. MARRIAGE AND BIRTHS HISTORY The next met of questions are for EVER-MARRIED WOMEN UP TO AGE 55 only. 'Now I am going to ask you questions MwmWT Y ikg l Thank alt the women oyer 55 yean for their time and explain that the rest o f the questions will be directed about the events of the last 7 years.' towards women with young children only. Begm collecting bformationfrom month of interview. 1. How old ate you now ? ______Mark 'X" at the month where the activity/ slate began and draw a line extending through ' 2 . M vihsXa%e did you mmry? (Include all marriages) _ the period into which the activity/state continued, then enter another 'X in the 3. Do you have any co-wivea? Y E S N O month that the activity ended M ly m i a If YÎES, how m any? ______and what rank are you?_ I 1. Water point function: 4. W hen was your last child bom ? _ m o n t h s .days (Ftw villages vrith water I development ONLY) S. A re you currently using any m ethod to stop yourself becom ing pregnant? Y E S N O Mark the months when the tap has been If YES, what m ethod? ______working over the last 7 years. From whom /where did you get it? ______I 2. Pattern of residence: Have you lived in this village 6. Have you ever used any m ethod o f contraception? Y E S N O continuously over the past 7 years7 If YES, what m ethod? ______Markali the months when you have lived In this village. 7. Are you still breast-feeding now? YES NO I 8. Did you menstruate this month? YES NO 3. Marital history: ? If NO, do you think diat you are prenant now? YES NO Have you been married continuously over the past 7 years? 9. BIRTH HISTORY Mark ail the months in which you have been married. N Year of Sex Father's name Was this child Is this child if dead, how birth .M/f adopted, or still alive? old was the 4. Husband residence: Have you lived with your husband I bom to Y/N child when it someone else? died? (yam continuously over the past 7 years? Y/N mdmomtu) Mark all ttte months your htaband has 1 been living with you

i BI - B6 Births The next six columns relate to all births 3 and their survival during the last 7 years. Each column re/êrs to a separate birth and its survival to time of interview, REFER to the completed birth history I § s For each birth (B1-B6) starting with the most 6 recent: & a) What month did you give birth? 7 Markthe month of birth X~ monthcf birth b) Did this child survive in each of the * following months (up to time of interview)? If the child died, tncàk the month of death X Draw a line connecting birth arid death. o I 7. sun births Did you have any other still births? S 1 Probe respondent for any stiO births, not Co 1 recorded in B1-B6. Mark any months with still birti Does this include all migrant sons? YES NO ?acar Does this include any children who died very soon after birth? YES NO Ï

Co The nest set o f questions are fo r E M W who liave EMW; HH: Village:. had a trirth during the LA ST 7 YEARS O N LY. Thank all other women for their tim e. MARRIAGE AND BIRTHS HISTORY Am m a waan waggaa torban darhan keessatU ta’e The >ezt set of questions are for EVER -M ARR IED W O M EN UP TO AG E 5.5 only. ilaalchiseea gaafii tokko tokko sigaafochn fedha. Mark ’X" ai the month where the acitvity/ 1. W aggaan kee meeqa? ______slate began and draw a line extending through I 2. Yeroo waggaan kee m eeqaatheenunte? Yoo yetoo tokkoo olbeerum tee jiiaatte kan heeiuma the period into which the iKtivity/state hundaa natti him ii. Iftaa ______2 fT a a ______3 M a a ______eontittued, then enter another 'X'in the I month that the activity ended 3. Abbaan w o ita keetii sim alee beeira fiiudhee jiiaa? A ti hoc isa afn iitii mceqafEaa? a Write nok/toul number of wives: _____ /______1. Water point function: (For villages with water I 4. Itm oon kee kan m aayyi yoom d h a ]a te ? _ . J » » _ development ONLY) Bonbaan kun yoom ganda kuoatti dhuabbate? Erga dhaabbate taklraa catwe YKN dhuabbate S. Ati amma akka hlndenyef wanti ati fiiyyadamtu nijiiaa? YES NO bcckaa mee yeroo itti dhuabate netti himi? Yoo jiiaate m aali? ______Eessaa argatte? ____ I 2. Pattern of residence: 6. tCnnmn eblrm tiin rimyrf wanti ati fownilatntee beektu iiraa? Y E S N O Waggaa torban darban hundtunsa kesessatti Yoo jiraate maxi foa fayyadanuiu turte ? 1 ______2 ___ 3 ganda kana keessa jiraatu turce? Mee yoomiina calqabee hanqa yoomi akka as jiraatte natti himtaa? 7. Ilm oo kee kan m ayyi ammalee haim a ni boosiftaa? Y E S N O Sokkaa ganda biraa deemte, YKN baati tokko I 8. Toibaanafianderbankeessattidugdiiattidhufejiraa? Y E S N O of asii tieemte bektaa? Yoo sitti dhufoe bin jiia a tin waan ulfboyte sitti nifokkaataa? Y E S N O D K 3. Marital history: 9. BIRTH HISTORY Mee waggaa torban darban kana keessati yeroo akka niitiitti YKN haadha wanatti itti jiiaatte M aqaa joolee keetii kan hundaa/kan o& i keeti dhaltee qofaa natti him ini dandeessaa? b/n Jachuun ilnuum kee kan bakka biraa jiraatan, kan Judhan/henwian akasumas kan sirraa natti himii? 4. Husband residence: [ Lak. Maqaa ibbaa dbaicfaee Yoodu'ê, yeroo aWsmlmW gssr4sAr*a—afk-m mmmm « m riio bmm/Wm Waggaa torban darban hundumaa keessatti abbaa W a g f u Y K N k M f ir V hibbuuo H (hi'e? wonaa/dtiirsa keeti in wajjiin mana tokko keessa galuu k e c tid ? Y /N nijiraaTY/N (yrt. and mths.) tuitanii? Yoomiraa hanqa yoomii akka dhiiraan 1 wajiin galuu rihaan dabarsite natti himuu dandeessaa? i Bl - B6 Births J. ... Irraagadeen jahan (6) itti aanan waa'ee hormaata waggaa torban dabranil/U lubbuun Jiraachuu t 4 Isaanll lUaala.Ibsa G-9ffa kan seenaa homusataatti fayldoomL § 3 ■ (fYaa ee du ’a ilmoo slgaqfaituun akka sigadbiisu I nanbeeka haatu'u malee akeeknl tptannoo kanaa 6 I tokko waa ‘ee kanaa beeku waan ta ‘eef facha 7 sigattfa chain dirgama natti taha.)

I a ) ____ J’ia kam keessa dhalate? Mcrk the month of birth‘X‘ 5 — b) akka du’e natti himte jiita kun ka ta*e 1» ji’a kam keessa? > 0 II 7. Still births Ibnoon sima deebi'e (kan du’ee dhalate) ni jiraa? S 12 Kan biroo hoo? Co TÏ------1 I ID number rr rri r Season: WET / DRY I I I I...J I Î Season: WET / DRY EMW:_ D a t e : ______EMW:_ D a t e : ______V i l l a g e : Interview er: ______V i lla g e : _ HH: HH: H EALTH SURVEY for ever-m arried women H E ALTH SURVEY for ever-m arried women (15-55) Now 1 w ould like to ask you some questions about the health o f you and your fam ily. Am m a gaafii w a’ee fayya keetifii kan m attii keetitin sigaafachuu barbaada. 1 1 . H o w o l d a r e y o u ? 1. U m riin kee meeqa? ______2. Are you still breast-feeding now? YES NO 2. Anuna harma nihoosiftaa: Ilmoo booda jechu kooti? YES NO 3. Do you think that you are pregnant im w ? Y E S N O 3. Amma ulfaaayeetin jira jatlee niyaada? YES NO 4. Have you A N D your children been immunised? (ask for immunisation cards to checl0 N Name Relationship to Vaccine/ Im m unisation Dates received N Maqaa Firummaa DHB B ifa talaali G tiyyaa itti fiidhatam e I E M W . e.j. AS" I* r* 3 " adopted SOI w ajiin qabu eg. §■ AS-adopted MO

S. D uring the past T W O weeks have you or any o f your children experienced ill health? 5. Totbaan lamaan darban keessattii ati Y K N ilm aan kee dhukubsatanii beektu? YES NO YESNO 6. Describe the sym ptom s and duration o f each period o f illness: 6. Mee m aal/akkam ittii akka si/ilm aan kee dhukubecG yeroo hangam if akka tahe natti him uu

N o Name Relationsh M Age Illness symptoms War Dtiratioo H o w d id y o u tre at Dawaa/ walaansa th e illness? N o Maqaa Firummaa S a l U m r B ifa dbukubicha^ i p t o o r d is e a s e o f illoea Dhukubti hanqamiif akkaro fiidhateAte ill that (days) 0 * DO tre atm en t DHB i n m u la t t o 0 is a a EMW dhukubsa gootaniif? F he/ihe 1 * tnditional chi w a j i i n M (use checklist) ciibsise O-wattakkallee n e eded to m e d ic in e p repared q a b u himwalaanomne e . ( . AS- lie etijiiaa? • d o p te d s o o 2 *■ tn d id o o a l F (days) d o w n ? e.g. AS " Y/N adaa/abashaa kan Y/N he ale r adopted soo manatti qopbayee 3 - h e a lth c lin ic 3-kiliiinikii

os Co

NO If YES, what are the sym ptom s_ Î Yoo sitti dhaga’ame akkam sigodha_ 0) §• CO Appendix III- VI: Questionnaires

Checklist of indigenous classification of illnesses (adapted from Slikkerveer (1990); Buschkens and Slikkerveer (1982)

English Oromo (indigenous cause and/or symptoms) amoebic dysentery albaati (diarrhoea) ascaria mukurtii tapeworm metoo (through eating raw meat) intestinal disorders mar’umaan gastritis singiggo (after eating chat) roundworm raammoo (after eating meat) fever lagda (umbrella term for fevers) nausea hogisa (through eating food which is off) influenza dufkaka (umbrella term for colds and flu) cough kufa whooping cough qakkee (children's disease) tuberculosis somba (caused by jiinii - spirit) pneumonia kuku/qilleensa (caused by jiinii) chicken pox guftii (children’s disease) malaria busaa (fever) anaemia alati back pain bukuu (through heavy work) migraine hurgunoo (sometimes caused by budaa) headache d’ukkuba mataa (sometimes caused by jiinii) muscular pain o’eekuu (caused by cold) stomach pain dukubi gara tooth ache dukubi like nose bleed funuuna (caused by jiinii) fracture lafee c’abe ear infection d’ukkuba gura trachoma d’ukkuba ijaa (disease of eyelids)

‘evil eye’ budaa (possession by an evil spirit) children’s disease shift! (often not regarded as a specific physical condition which may include measles, chicken pox, whooping cough, dysentery and German measles)

Health codes and values; 01 = cold 06 = other stomach cramps 11 =gyn/obstet. 02 = fever 07 = headache 12 = anaemia 03 = respiratory infection 08 = back/muscle pain 13 = malaria 04 = diarrhoea & gastritis 09 = eye infection 14 = other 05 = intestinal parasites 10 = skin infection ?

EVER MARRIED WOMEN (EMW) QUESTIONNAIRE EVER MARRIED WOMEN (EMW) QUESTIONNAIRE I EMW:______H H :______EM W :______HH: Interviewer. Interviewer. ? Villagc:_ Date: Village:_ Date: 5 Theee question: ihonM be addreaaed to all ever-married women within each hongehold Amma waa’ee biahaan mana kecaatti dhngaatiifl waan adda addaatiif itti fayadoomtan (D HOUSEHOLD ACCESS TO WATER In the DRY season In the WET season ilaalchiien ligaatachnu barbaada. QUESTIONS HOUSEHOLD ACCESS TO WATER In the DRY season In the WET season 10. What is your main source of drinking TAP / SPRING/ TAP / SPRING/ QUESTIONS water? (circle answer) RIVER/ RIVER/POND RIVER/ RIVER/ 10. Bishaan dhugaatii, nyaataan Maagada /Bombaa /Eela Maagada /Bombaa /Eela /SPRING/ POND/SPRING/ bilcheeifachuu fi waan tokko tokkoof itti /Lags/Haroo/Madda /Laga /Haroo /Madda / 1 RAINWATER/ RAINWATER/ faadoomtan baayyisee kan argattan Kan biraa Kan biraa OTHER OTHER maalirraahi /eechairahi?(circle answer) (Birkaaraa yoo ta 'e ) If answers TAP ask: a) Biikaa haara’a WaterAction YKN Y N Y N a) Do you use the new tap stand installed in YES NO YES NO WaterAid ____ itti dhaabbe inaa ni this village by Water Action? waraabbattuu? b) Biikaa biraa kan ganda biraa Y N I Y N b) Do you use any other tap stands or pumps YES NO YES NO keessaa inaa ni waraabbattuu? in the surrounding villages? Yoo waraabbatan ganda kam? 11.__ kana bona bona/ ganna ganna erga Years Years If YES, in which villages? fayadamu calqabdani bara meeqa.Ji'a Months Months kamkeessa? 11. How long have you been using this ...... Years Years 12. Mana keeti hanqa dhayxu Hours Hours season water source? hanqam sinaa fudhata? (x 2) Minutes Minutes Months Months 13. kana yeroo dhayxe booda bishaan Hours Hours 12. How long does it take for a single trip Hours Hours waraabbachuuf hanqam nama egsisa? Minutes Minutes from your home to the water source? Minutes Minutes 14. Jeerikaani moo hubboo dhaan wanabda? J I J I 13. How long do you wait in a queue at the Hours Hours the source? 15. Dhaqa tokkotti bishan litri meeqa Litres Litres Minutes Minutes waraabbatta? t 14. Do you use a jcrry-can or insera? (circle) J I J I 16. Ati qofhin bishaan kan waraabdu torbaanitti yeroo meqaa? Trips Trips § 15. How many litres are collected on a single Litres Litres 17. Gaangee .harree ,YKN fi«rH«iinbishaan trip to the water source? ni waraa bbattaa ? Y N Y N & 16. How many trips to collect water are m«d<» Yoo warraabbatte torbaanitti dhaqa meeqa per week? Trins Trins beeyladaa Ifanaan waraabbatta? 18. Yeroo waraabachuuf dhaqxu namni 17. Do you use either a mule or donkey? YES NO YES NO bishaan siif ba'atuun sigargaaru nijiraa? Y N Y N Ifso, how often? Yoo jiraate maal siif ta’a? Torbaanitti dhaqa 0 18. Does anyone assist you by carrying the meeqa siearsaara? R/N R/N water for you? YES NO YES NO 19 Namni kiia’a kafalte kan waraabsiflu § nijiraa? Y N Y N Co If YES, then wtm does? (state name and 20. Bishaanin waraabbachuuf meeshaa kan relation to EMW) R/N R/N akka, garii YKN makiinaadhaa Y N Y N 19. Do you pay someone else to collect and nifayadarataa? cany water? YES NO YES NO 1 20. Do you use a cart or car for water collection? YES NO YES NO I Appendix VII: Event history calendars

Use of event history calendars

The calendar employed recorded information on births and survival outcomes, as well variables that might have influenced timing of these events e.g. periods of water point function, residence patterns and marital status. These life history events were dated to the year and the month using a reference calendar of local events, including historical events such as the water point installation, religious festivals (like Orthodox Christmas, Eid-ul Fitr), political events, and agricultural seasons.

The calendar format was a large grid (Appendix V). One dimension of the matrix detailed the behavioural pattern/event being investigated; the other dimension was divided into time units for which these events were recorded. The interviewer filled in the cells of the matrix with information provided by the respondent, recording his/her status across the time intervals.

Events occurring in each domain were recorded within one colunrn on the calendar as a series of dichotomous variables: a) whether or not the water point was functioning; b) whether or not the respondent was living in the study village; c) whether or not the respondent was married; d) whether or not the husband was living with the respondent. For the final domain concerning birth and infant survival, each birth recorded during the study period was entered on a separate column (B1 - B6). After establishing a birth the enumerator was required to probe the respondent as to the survival status of that child in the subsequent months and record any mortality event in the same column. A further column (7) recorded the months of any still births or miscarriages occurring during the entire observation period.

The timing of data was recorded by marking the beginning and ending months, thus defining a ‘spell of activity’, between significant events. An “X” was entered in the month that this activity began e.g. date of birth; then a line drawn extending through the time period in which this state continued, and another “X” was entered in the month where it ended, e.g. date of death. To record the timing of a number of events within a column, e.g. periods of husband residence, the enumerator started with the most recent event to the time of interview recording the timing of events backwards through time. Appendix VII: Event history calendars

Strengths • It can improve the quality of retrospective data by helping the respondent to relate both visually and mentally, the timing of several kinds of events. • Events more readily remembered, such as births, provide important reference points for recalling less salient events such as living arrangements. • The calendar display calls to the attention of both respondents and interviewers any inconsistencies in the timing of events in different domains, at the time of enumeration. • Detailed sequences of events are easier to record with an EHC than with a questionnaire, e.g. recording numerous changes in residents patterns over the years. • Previous studies comparing calendar and tabular formats have indicated that the calendar estimates are closer to those from current status census data (Freedman et al., 1988). • An individual may experience several events over the observation period with the number of events varying considerably across individuals, which can be recorded on the EHC.

Limitations • Necessitates intensive interviewer training. • Coding data may be time consuming, when involving fairly detailed entries with a large number of data points.

Event histories typically possess two features, censoring and time-varying explanatory variables, that create major problems for standard statistical procedures such as multiple regressions, which may result in severe bias, or loss of information - however, a number of statistical methods have been developed to deal with this kind of duration data.

Event history analysis Potential problems initial selection bias - differences in the effects of co-variates on hazards rates controlling for time may reflect differences among groups of people found at different co-variates at each time, rather than the real effects of states on hazards rates; however with life event histories it is not possible to select a random set of people in different Appendix VII: Event history calendars states which would eliminate selection bias (Allison, 1984). unobserved heterogeneity - omission of variables that influence both transition in and out of co-variate states and hazard rate of events under study can lead an overestimation of the negative duration effect or positive duration effect (Wood, 1994). This can sometimes be addressed using multi-level modelling. reverse causation - referring to the influence of the dependent process on the co-variate process. This becomes an issue because the effects of the co-variate on the hazard rate will be confounded with effects of the dependent process on the value of the co-variate. Appendix VHI: Training manuals

UCL Training Manual ■' ' UNIVERSITY COLLEGE LONDON Mhairi Gibson, Dept, o f Anthropology October, 1999 What is this study ahout? This is a study of demography, health and nutrition. The focus of this study is to investigate the demographic, health and nutritional impact of a recent water development project in rural Arsi.

What is demography? Demography is the study of people. It looks at: • fertility - the number of children a woman has given birth to • mortality - the death of people (adults and children) • migration - the movements people make

Why are we doing this study? Many development projects aim to improve welfare and alleviate poverty of rural women by reducing the time and energy expended on household tasks. This study is the first comprehensive investigation of the demographic changes that may occur following water development in rural areas.

In rural Ethiopia women can spend up to 6 hours a day collecting water. One of the primary objectives of water development is to make clean water available at reasonable distances, thus reducing the physical stress on women and the time spent in water collection. Since 1993 WaterAid (a UK NGO) and Water Action (an Ethiopian NOG) have introduced a water development scheme designed specifically be bring water supplies closer to a number villages in the Hitosa wereda. Fertility and mortality levels may be effected by these developments.

It is important to examine the demographic changes that have occurred following the introduction of these development technologies since the effects of population growth, in a region already characterised by high rates, may present the most serious social problems of the next century.

This study has been designed to with two general aims: • to examine the both the changes for women within these villages over time, before and after water development. • to compare these villages with neighbouring ones currently without water development

Research status The researcher (Mhairi Gibson) is from University College London (UK) and is affiliated with the DTRC, IDR, Addis Ababa University. She has obtained all the necessary formal Appendix VIII: Training manuals

documents from the regional authorities and has permission to carry out this study from WaterAid and Water Action.

Introduction It is important to tell people the following information • the data collected is confidential • it will not be passed on to the government /other officials

What will the respondent get out of the study? This is a question which is often asked by people when they are asked for information. It is a very difficult question to answer. This information has never been collected before, which is why we are collecting it. We hope that when this information has been collected, it can be used by development groups to inform decisions about factors such as health care and water provisioning.

It is very important that you do not promise people something that is not true. For example, do not say that as a result of this information, a clinic or more water taps will be built in the area.

Questionnaires For this project there are 3 questionnaires:

Head of Household (HH) Questionnaire The unit of study for this questionnaire is the household. The household is defined as a man, his wives, unmarried children and any other people who live with them (e.g. old relative). Each household will have a household questionnaire and should be completed by the head of household.

Ever-Married Women’s Questionnaire The unit of study for this questionnaire is all ever-married women. In each household all ever-married women should each complete this questionnaire. This questionnaire is composed of three sections relating to: • marriages and births • a recent events calendar • water access All three sections should be completed separately for each ever-married women.

Health and Nutritional Survey The unit of study for this questionnaire is a sample of the ever-married women who are reproductive-aged (15 -55 years). One questionnaire will be completed for each women. Anthropometric measurements for those women who are NOT pregnant will then be taken by the researcher, for weight, height and skinfold thickness. Appendix VIII: Training manuals

Get permission When arriving at a household, speak with the head of the household, to explain what we are doing. Do not go ahead and attempt to interview people without permission from either the head of the household or another elder. Remember to ask an adult male if you may speak with his wives.

Before beginning the interview, complete both the identification information at the top of the paper: your name village name the date and time name of household head name of respondent ethnic group AND religion of household Examples of ethnic groups: Oromo, Amhara, Afar Examples of religion: Muslim, Orthodox, Traditional, Protestant

At the very top of the paper is an idenfication code box, which will be completed by the researcher after the interview.

The six-digit code refers to: first digit = village number (1-8); next three digits = household number (001-300); final digit = ever-married woman number (1-4) eg. the third women interviewed in household number 121 in village number 2 =21213

Completing the questionnaire sheet

General notes • when the response is zero, write 0 • when the person being questioned does not know the answer, write DK (DK = Do not Know) • do not leave a square empty if it should have some information in it • write any extra information on the back of the sheet Appendix VIII: Training manuals

Head of Household Questionnaire

Who to ask It is unlikely that any one person will be able to supply all the information. Start with the household head (usually the man). If he does not want to answer the questions, or is unable to give all the information, ask if you may speak with his wives.

For Questions 1 to 4, work DOWN the sheet. For the following questions, work ACROSS the sheet.

(1) Name “Please give me the names of all the persons who usually live in your household, including people who are not actually here at the moment and anyone who stayed here last night. Start with the head of household”

We want information on every person in the household, including old people, young babies, people who are away on business etc. If there is a young baby who has not been named, write “baby” Do not include children who now live permanently elsewhere e.g. if there is a daughter who is married and has moved to live with her husband, then she is no longer part of the household we are interested in.

(2) Sex “I s male or female?”

What is the sex of each person. Do not guess from the name. Write M = male F = female

(3) Relationship “What is the closest relationship o f in the household?

For each person listed in (1), find out what is their closest relationship to the head of household. This is done in terms of line number. Use the following codes: Head of household HH Husband H Wife W Daughter D Son S Adopted child A Mother M Father F Brother B Sister S Appendix VIII: Training manuals

Obviously, a person will have several relationships in one place. For example, a young boy will be the son of his mother and father, the brother of his brothers and sisters etc. etc. What we are interested in is his closest relationship, so we would complete that he is the son of his mother and father.

Example: John is the head of household. Sarah is his wife. Jane is the daughter of John and Sarah. Peter is the adopted child of Sarah.

Number Name Sex Relationship (1) (2) (3) 1 John M HH 2 Sarah F W o f 1 3 Jane F D of 1 + 2 4 Peter M AS of 2

(4) Age “How old is ?”

It is necessary to obtain each person’s age as precisely as possible. You cannot guess from just looking at them. If an individual does not know their age, there are several things you can do.

1. If a person knows the year in which they were bom, you can work out their age using the calendar 2. You can ask if a person is older or younger than someone who does know their age, and ask what is the difference in their ages. 3. You can ask if the individual was bom before or after a specific event, or if there was something that happened in the year that they were bom, then use the calendar to work out their age.

Try to check that the ages given are reasonable. For example, if there is a mother and her son listed, and there is less than 15 years difference in their ages, don’t guess, but say that perhaps you have made a mistake, and try to sort out the problem by asking people about their relative ages.

(5) Mother’s survival “I s birth mother still alive?”

Mother refers to the woman who gave birth to that person, not necessarily the person they grew up with. Write: Yes V No N Appendix VIII: Training manuals

Do not know DK

(6) Father’s survival “ I s birth father still alive?”

Father refers to the man who helped the birth mother to give birth to that person, not necessarily the person that they grew up with. Write: Yes Y No N Do not know DK

(7) Education “How many years of education has completed?”

Write the number of years education completed in government schools and religious schools separately. If the person says that they have never attended school, ask if they have ever attended adult education or any religious schools.

If a person has never had any formal education write 0 If the answer is not known, write DK

(8) Residence patterns “How long has lived in this village continuously?”

Write the number of years and months that the person has lived continuously in the village. This should include any periods of temporary absence for business trips, herding etc.

If the person has been living in the village since birth, write A (= always) If the person has been living in the village less than 2 years, record in months

(9) Movement “What w as main reason for moving to this village”

For each person find out the original reason why they moved to the village. If the person has been living in the village since birth, write A (= always)

Use the following codes: Government resettlement GR For marriage M Forjob J For better land L To receive inheritance I For better access to water W For better access to health clinic HC Appendix VIII: Training manuals

If the person mentions any other reason, write the fiill reason in the space or in the space at the top right side of the paper and make a note of the line number.

(10) Marital status “At the moment, is married/ widowed/ separated/ never married?”

Married M Widowed W Separated S Divorced D Never married NM Do not know DK

Marriage is taken as when somebody has lived with another person as man and wife.

(11) Age at first marriage “At what age did the person FIRST live with somebody as man and wife?”

Men who have more than one wife - this question is asking about their FIRST wife. Women who have married more than once - this question is asking about their FIRST husband, even if he is now dead.

If the respondent does not know the exact age, use the same technique as before when finding out about the person’s current age.

The following set of questions are only to be asked for women who have ever been married. In column (10) they should have recorded either married or divorced or widowed or separated. This question should not be asked for women who have never been married.

(12) Wife rank and number of co-wives “What number wife is the woman, and how many co-wives does she have at the moment?”

Write rank/ total number of wives The total number of wives will include the respondent.

Example: The women is the third wife and has 3 co-wives. You would write 3/4

Example: Appendix VIII: Training manuals

The women is the only wife. You would write 1/1

This question can be used to check the information about household residents (1). For example, if a man says that he currently has four wives, but only 3 are mentioned in column (1) as household members, then ask where the other wife lives. Perhaps she lives outside the district. Make a note of this information on the back of the questionnaire.

Questions (13) to (19) are asking about children the women actually gave birth to. We are not interested in children she is looking after for someone else, or who have been adopted by her.

(13) Number of sons currently living at home “How many sons, that ...... actually gave birth to, are currently living in this household?”

Write the total If the answer is zero, write 0 Go back and check column (2) to make sure of this reply. Query any discrepancy.

(14) Number of daughters living at home “How many daughters, that actually gave birth to, are currently living in this household?”

Write the total If the answer is zero, write 0 Go back and check column (2) to make sure of this reply. Query any discrepancy.

(15) Number of sons living elsewhere “How many sons, that actually gave birth to, are alive but living elsewhere?”

Write the total If the answer is zero, write 0

(16) Number of daughters living elsewhere “How many daughters, that ...... actually gave birth to, are alive but living elsewhere?”

Write the total If the answer is zero, write 0

Sometimes if happens that children die. If may be very sad for you to talk about and I am sorry to ask you about painful memories, but it is important to get the right information. Appendix VIII: Training manuals

(17) Number of sons dead “H as...... ever given birth to a boy who was born alive, but later died?”

It does not matter at what age they died - it could have been as a very small baby or its could have been when they were grown up. Write the total If the answer is zero, write 0.

(18) Number of daughters dead “H as ever given birth to a girl who was born alive, but later died?”

It does not matter at what age they died - it could have been as a very small baby or its could have been when they were grown up. Write the total If the answer is zero, write 0.

(19) Check “Just to make sure that I have this right -...... has had a total births during their lifetime. Is that correct?”

This is a question for you to check that the respondent has not forgotten any children. Add the numbers in columns (13) to (18) and ask whether this is the total number of children the woman has ever given birth to. If the respondent disagrees with the total you suggest, then go through each column again until the totals balance out.

(20) Economic activity of household “What is the main economic activity of this household”

We are concerned with the way in which the person earns a living. The main activity should be in terms of the activity in which most time is spent. Try to be specific as possible. Examples of answers: • cultivation (indicate the main crop e.g. maize, potatoes, beans) • agro-pastoralism (pastoralism and cultivation) • shop keeper • teacher

(21) Livestock “Does any of the people you have listed as a member of this household own any livestock?”

We are concerned with the variety and number of livestock owned by members of the household. It is important to record the number of each variety as well as who owns them. Write down the line number of the person who owns the livestock Appendix VIII: Training manuals

(22) Possessions *Does any of the people you have listed as a member of this household own any of the following things?”

For each item listed, ask whether any member of the household listed above owns one of them.

(23) Sanitation “What kind of toilet facility does your household have?”

Circle the relevant method of waste disposal. Note down any other methods of waste disposal that the respondent may mention.

REMEMBER: Make a note of anything interesting or relevant that comes up during the interview, but which is not recorded on the questionnaire itself. Make notes on the back of the sheet.

Thank the respondent for his/her time.

Debriefing: Each questionnaire sheet must be checked and signed by the researcher after the interview. It is important to discuss your impression on the quality of the interview. Did it go well? Did you feel that people were being truthful? Appendix VIII: Training manuals

UCL UNIVERSITY COLLEGE LONDON Training IManual Supplement Mhairi Gibson, Dept, o f Anthropology Ever-Married Women’s (EMW) Questionnaire October, 1999

Who to ask The unit of study for this questionnaire is all ever-married women up to the age of 55 only.Explain to all women over that age that this questionnaire will be directed towards younger women with young children.

One questionnaire should be completed with EACH ever-married woman listed in the house. This questionnaire is composed of three sections relating to: • marriages and births • a recent events calendar • water point access All three sections should be completed separately for each ever-married women up to 55 years of age. Check with column (10) in the household questionnaire that all these ever- married women have been included.

Before you start, remember to fill in ALL the identification details (EMW name, HH name and village name, your name, date and time)

(1) Age “How old are you now?”

It is necessary to obtain each person’s age as precisely as possible. You cannot guess from just looking at them. If an individual does not know their age, there are several things you can do.

1. If a person knows the year in which they were bom, you can work out their age using the calendar 2. You can ask if a person is older or younger than someone who does know their age, and ask what is the difference in their ages. 3. You can ask if the individual was bom before or after a specific event, or if there was something that happened in the year that they were bom, then use the calendar to work out their age.

Try to check that the ages given are reasonable.

(2) Age at marriage “At what ages did you marry?” Write the age that the woman started her married life. If the woman has been married more than once, write her age at the start of EACH marriage. Appendix VIII: Training manuals

(3) Co-wives “Do you have any co-wives?” The total number of wives will include the respondent. Record her rank.

Example: The women is the third wife and has 3 co-wives. You would write 3/4 Example: The women is the only wife. You would write 1/1

This question can be used to check the information about household residents (1) in the Household questionnaire. For example, if a woman says that she is currently one of four wives, but only 3 are mentioned in column (1) as household members, then ask where the other wife lives. Perhaps she lives outside the district. Make a note of this information on the back of the questionnaire.

All the following questions are asking about children the women actually gave birth to. We are not interested in children she is looking after for someone else, or who have been adopted by her.

(4) Most recent birth “When was your most recent child born?” Write in years and months If the child was bom less than 2 years ago, use months. If the child was bom less than 1 month ago, use days.

(5) Current contraceptive use “Are you using any method to stop yourself getting pregnant now?”

Write the method and source of contraception that the woman mentions she is using. Examples of Methods: Examples of Sources: Periodic abstinence Government Hospital Withdrawal Local health clinic (state village name) Breastfeeding Shop Condom Pharmacy The Pill Community health educator Intra-Uterine Device (lUD) Traditional healer Traditional methods (state the name) Friend Sterilisation

(6) Past contraceptive use “Have you ever used any method to stop yourself getting pregnant?” Appendix VIII: Training manuals

Write all the methods of contraception that the women has used in the past.

(7) Breastfeeding status “Are you still breastfeeding your most recent child?” Record the women’s current breastfeeding status. She may still be breastfeeding her older children, even if she has not given birth recently.

(8) Menstruation “Did you menstruate this month?”

We are concerned with whether she has experienced menstrual bleeding during the last 4 weeks. Check that she has had a full menstrual cycle, if she has recently given birth then it may only be post-partum bleeding. If the woman has NOT menstruated in the last 4 weeks. Ask her whether she thinks that she may he pregnant. “Do you think that you are pregnant now?”

(9) Birth History “Please list all the children that you have ever given birth to”

Start by writing all the names of the children in the first column. Then work across the grid for each child for each of the following questions; • “What was year of birth?” Use local calendar. If the respondent does not know, ask the age of the child then calculate the year. • “What is sex?” Write M or F • “What is the name o f biological father” We are concerned with the man who made the woman pregnant with the child, do not assume it is the man she is married to now, or that all children are fathered by the same man. • “I s adopted/fostered or originally born to someone else?”

Sometimes if happens that children die. If may be very sad for you to talk about and I am sorry to ask you about painful memories, but it is important to get the right information.

• “I s...... still alive?” • “How old was the child when it died?” Write years and months. If the child was less than 2 years, write in months.

Use the following questions to CHECK that ALL children have been included: Does this include all married daughters? Appendix VIII: Training manuals

Does this include all migrant sons? Does this include any children who died very soon after giving birth? Go back and enter them in the birth history. SEVEN YEAR EVENTS CALENDAR The next section of the questionnaire is concerned with the events of the past 7 years. It should be completed ONLY by ever-married women who have experienced a BIRTH in this time. Check the birth history (9) to see whether the woman has had a birth in this time.

For ever-married women who have NOT experienced a birth DURING THE LAST 7 YEARS go straight to the ACCESS TO WATER SECTION (QUESTION 10)

What is an events calendar? An events calendar is a method for collecting information over a period of time. In this case we are concerned with the period of the last 7 years. On the left side is a list of the months for this time period starting with the month of interview at the top of the page. On the calendar we can record the distribution of events and activities over this time period. Each column represents a different type of event or activity.

Begin collecting information from the month of interview, and work from the top to the bottom of each column. Write X at the month where the activity began and draw a line extending through that period into which that activity continued, then enter another X in the month that the activity ended.

“ Now I going to ask you questions about the events of the last 7 years.”

(Column 1): Water point function FOR VILLAGES WITH WATER DEVELOPMENT ONLY “Has the water point which you use most often been working continuously over the past 7 years?”

Mark all the months when the tap has been working over the last 7 years. You may show the respondent the calendar to help them recall. Write X at the month that the water point was installed, then draw a line extending through the months where the tap continued to function, then enter another X in the month that the water point was broken or not functioning. If the water point was fixed record the period that it continued to function in the same way.

(Column 2): Pattern of residence “Have you lived in this village continuously over the past 7 years?

In the same way mark all the months that the woman has been living in this village. Starting with the month of interview record all the months she has been resident. (Column 3): Marital history “Have you been married continuously over the past 7 years? Appendix VIII: Training manuals

Mark all the months in which the woman has been married starting with the month of interview. Record the start and end of each period of marriage, if she has been married more than once.

(Column 4): Husband residence “Have you lived with your husband continuously over the past 7 years?

Mark all the months that the woman has been living with any man/husband. Record the start and the end of each period of cohabitation. Do not attempt to record any periods of absence less than one month.

(Column Bl-7): All births The next 7 columns (Bl-7) relate to all births and the survival of each of these births during the last 7 years.

EACH column refers to a separate birth and that child’s survival up to the time of interview.

1. Begin by asking the woman how many children she has given birth to over the last 7 years, then you will know how many columns to use.

For example: Mary has a baby of one month, a 2 year old child and a daughter of 15. So she has had 2 births in the last 7 years and 2 columns should be completed.

2. CHECK the completed birth history (9) that all births OVER THE LAST SEVEN YEARS are included.

• Month of birth - For each birth mark the month of birth with an‘X’ ‘What month did you give birth to ’

• Month of death - Record whether each child survived in each of the following months up to the time of interview. Ask the women about each month from the birth. If the child died, mark the month of death with another ‘X’ ‘Did this child continue to live up to the present day?....in which month did it die?’

3. Draw a line connecting birth and death.

REMEMBER child death is a sensitive subject. It is important to treat this subject matter sensitively, however it is important to get as detailed information as possible. If you already have accurate information about the number of children who have died during this time period, then don’t repeat the question, simply gently ask for more detailed timing on the month of death. ACCESS TO WATER ‘Now I would like to ask you some questions about your water supplies’ Appendix VIII: Training manuals

The final section of the EMW questionnaire is on a separate sheet. These questions are about household water collection.

There are ten questions which must be answered for each season. There are two columns, one for the DRY season and one for the WET season Starting with the DRY SEASON, work down each column answering each of the questions. Then do the same for the WET SEASON.

(10) Main source of water What is your main source of drinking water?’

The main source is one used most frequently, contributing the largest amount of water to the household. Circle the answer from the list. If another answer is given, circle OTHER and write it in the space provided.

If the answer is TAP or PUMP, ask the following questions a) Do you use the new tap stand installed in this village by Water Action?’ b) ‘Do you use any other taps or pumps in any villages nearby?’ If YES ask ‘ In which villages?’

Write the names of these villages.

(11) Length of use ‘ How long have you been using this DRY/WET season water source?’

Write the length of time in years and months.

(12) Total collection time ‘ How long does it take for you for the entire trip, to go there, collect water and come back?’

Write the length of time in hours and minutes.

(13) Waiting time ‘ How long do you usually wait in line to take your turn to get water at the ...... 9 »

Write the length of time in minutes. We are interested to learn how long women must wait in line to collect water once they arrive at the water source.

(14) Volume of water Appendix VIII: Training manuals

‘ How many litres are collected on each trip to the water source?’

If the woman does not know, ask the size and number of jerry cans/ insera. Calculate the total amount in litres. For example: two 10 litre jerry cans and one 15 litre insera = a total of 35 litres

(15) Trips to collect water ‘ How many trips to collect water are made per day?’

Record the frequency of trips to go and collect water. If less than one trip is taken per day write the number of trips per week. REMEMBER to write down whether you are recording trips PER DAY or trips PER WEEK.

(16) Women carrying water Do you go to collect water and return carrying it yourself?’

We are interested to find out if the respondent carries the water herself.

(17) Water carrying helpers Does anyone assist you by carrying the water for you?’

We are interested to know if anyone else in her family or elsewhere carries the load for her.

If YES, list the names as well as the relationship of these people to the women. For example: Mesret’s daughter Maryam and young son, Yohannis help her. Write: Maryam (D) + Yohannis (S)

(18) Use of professional water carriers ‘ Do you employ someone else to collect and carry water?’

(19) Use of donkeys for water collection ‘ Do you use donkeys for water collection?’

We are interested to learn whether the woman uses donkeys to carry the water load. And how often she uses the donkeys in terms of number of trips. If YES, ‘ how often?’ Circle the answer from the list: all trips (100%) most trips (75%) half of all trips (50%) some trips (25%) very rarely (less than 25%) (20) Use of vehicles for water collection ‘ Do you use carts or other means of vehicle transport to collect water?’ Appendix VIII: Training manuals

We are interested to learn whether the woman uses any form of vehicle or cart to assist in water collection.

FINALLY REMEMBER: Make a note of anything interesting or relevant that comes up during the interview, but which is not recorded on the questionnaire itself. Make notes on the back of the sheet.

Thank the respondent for his/her time.

Debriefing: Each questionnaire sheet must be checked and signed by the researcher after the interview. It is important to discuss your impression on the quality of the interview. Did it go well? Did you feel that people were being truthful? Appendix VIII: Training manuals

UCL UNIVERSITY COLLEGE LONDON Mhairi Gibson, Dept, o f Anthropology October, 1999 Training Manual Supplement Ever-Married Women’s Health and Nutrition Survey

Who to ask The unit of study for this questionnaire is all ever-married women up to the age of 55 only. So this means all the women who completed the ever-married women’s questionnaire.

One questionnaire should be completed with EACH ever-married woman under 55 listed in the house. The survey is composed of two parts relating to: • health status of the woman and her children • nutritional status of the woman

Two phases of data collection Each women interviewed will be surveyed twice. Once during the wet season and then repeated during the dry season.

For this reason it is important to accurately record the identification information for each women when you visit a household for the first time, so that you can find her again and repeat the questionnaire during the next season.

Before you start, remember to fill in ALL the identification details (EMW name, HH name and village name, your name, date and time) Circle the season in which the interview takes place.

‘Now I would like to ask you some questions about the health of you and your children’

(1) Age “How old are you now?”

IF this is the first round of questioning, then you have this information already. There is no need to repeat the question. Simply find the information on the EMW sheet question (1) and fill in the answer.

HOWEVER, IF this is the second round of questioning ask the question again. It is necessary to obtain each person’s age as precisely as possible. Appendix VIII: Training manuals

(2) Breastfeeding status Are you still breastfeeding your most recent child?”

IF this is the first round of questioning, then you have this information already. Simply check question (7) on the EMW sheet and fill in the answer.

HOWEVER, IF this is the second round of questioning ask the question again. Record the women’s current breastfeeding status. She may still be breastfeeding her older children, even if she has not given birth recently.

(3) Pregnancy “Do you think that you are pregnant now?”

IF this is the first round of questioning, then you have this information already. Check question (8) on the EMW sheet and fill in the answer.

HOWEVER, IF this is the second round of questioning ask the question again.

(4) Vaccinations/Immunisations “Have you and your children been immunised”

• First, copy the names of all children recorded in the birth history in question (9) on the EMW sheet • Write the relation of each child to the woman e.g. D = daughter, AS = adopted son

• Some families may have immunisation cards. Ask to see immunisation card for BOTH the women and ALL her children. Copy the vaccination dates for each vaccine from each card

• Start with the woman herself, asking the name and dates of each vaccination Then complete the information for each child. Write the month and year of each vaccination

• IF the woman has no cards, ask her to recall the information using the following prompts:

Vaccinations that they may have had: BCG - injection in the arm, or shoulder causing a scar Polio - drops in the mouth which may have been received up to 4 times. Record the date of each Polio 0 Polio 1 Polio 2 Polio 3 DPT - injection in the thigh or buttocks which may have been received up to 3 times Appendix VIII: Training manuals

Record the date of each DPT 1 DPT 2 DPT 3 Measles - injection

(5) Illness recall “During the past TWO weeks have you or any of your children experienced ill health”

(6) Health diary This table should be used to describe all the symptoms and duration of each period of illness that the women and all her children have experienced DURING THE PAST 2 WEEKS. Use the checklist provided to help the woman remember any periods of illness and code the response.

Fill in the woman’s name and age. • Ask her for the name of any illnesses she has experienced during this time. For each ask her: • “Were you so ill that you needed to lie down and remain in bed?” • “ How many days were you ill” • “ How did you treat this illness”Use the codes provided

Fill in each child’s name, relationship to the woman, sex and age from the information you have collected already. Repeat these questions FOR EACH CHILD.

Health codes and values: Treatment codes and values: 01 = diarrhoea and gastritis 00 = no treatment 02 = fever 01 = traditional medicine prepared 03 = respiratory infection at home (breathe faster with shorter breaths) 02 = traditional healer 04 = cough 03 = health clinic (write the name 05 = cold, flu, runny nose the village where the clinic is 06 = headache situated) 07 = back ache, muscular ache 08 = eye infection 09 = night blindness (vitamin A deficiency) 10 = children’s disease 11 = unknown 12 = other...... (write the name) 00 = none Appendix VHI: Training manuals

(7) Current health “Do you feel ill today?”

We are interested to learn whether the woman is feeling well on the day of interview. IF she says YES, she does feel ill, describe the symptoms using the codes.

NEXT.... ANTHROPOMETRIC MEASUREMENTS.

Who to ask The nutritional survey, requires a number of body measurements to be made for all ever- married women who are NOT pregnant.

Check question (3) of the health survey to find out whether the woman is pregnant.

IF she is not, explain the measuring procedure to her and ask whether she will agree to some quick and simple measurements which will determine her health.

“ Now with your permission I would like to take some measurements of you body”

The researcher (Mhairi Gibson) will take all measurements of: height weight arm circumference 2 measures of skinfold thickness

These measurements will take place within the privacy one hut selected in the village. The measurements of arm circumference and skinfold thickness will require each woman to remove some of her upper body clothing.

FINALLY REMEMBER: Make a note of anything interesting or relevant that comes up during the interview, but which is not recorded on the questionnaire itself. Make notes on the back of the sheet.

Thank the respondent for his/her time.

Debriefing: Each questionnaire sheet must be checked and signed by the researcher after the interview. It is important to discuss your impression on the quality of the interview. Did it go well? Did you feel that people were being truthful?