E'IIIII'I'IIIn

'l' fibula

3..»

This is to certify that the
dissertation entitled

THE ECONOMIC IMPACT OF PRIME-AGE ADULT
MORTALITY ON MALAWIAN AGRICULTURAL
HOUSEHOLDS IN THE ERA OF HIV/AIDS

presented by

EDWARD PEPUKAYI MAZHANGARA

has been accepted towards fulfillment
of the requirements for the

degree In Aricultpral Economics

 

 

 

2W 3%

Major Professor’s Signature
Aug, 2G1 'ZCO7

Date

 

MSU is an aflinnative—action, equal-opportunity employer

 

LIBRARY
Michigan State
University

 

o-I-u-o-u-o-u-u—-----o-c-.-o-o—.—o—u—o-o-o—o—u—c----—u-o—-—----u-u—u-o-o-u—o-n---v—y—o-c-o-0—.-.—-—-.-—--—---u—.—u—c-u.

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATEDUE

DATEDUE

DAIEDUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/07 p:/ClRC/DaleDue.indd-p_1

..__ -Hw__a_... _ s_‘

 

THE ECONOMIC IMPACT OF PRIME-AGE ADULT MORTALITY
0N MALAWIAN AGRICULTURAL HOUSEHOLDS
IN THE ERA OF HIV/AIDS

By

Edward Pepukayi Mazhangara

A DISSERTATION

Submitted to
Michigan State University
In partial fulfillment of the requirement
for the degree of

DOCTOR OF PHILOSOPHY

Department of Agricultural Economics

2007

ABSTRACT

THE ECONOMIC IMPACT OF PRIME-AGE ADULT MORTALITY
ON MALAWIAN AGRICULTURAL HOUSEHOLDS
IN THE ERA OF HIV/AIDS .

By

Edward Pepukayi Mazhangara

Prime age adult mortality has been increasing with the advent of the Human
Immunodeficiency Virus/ Acquired Immune Deficiency Syndrome (HIV/AIDS)
pandemic. Sickness, funerals and post funeral ceremonies rob households of human and
financial capital, which has potential to harm the welfare of agricultural households in the
short, medium and long term. This study examines the impact of prime—age adult
mortality (PAM) on agricultural production and rural non-farm employment participation
among farm households in Malawi.

The first essay uses a difference-in-differences fixed effects regression to analyze
a 13-year interval panel data set to examine the effects of PAM in the short (O-3years),
medium (4-6 years) and long term (7-13years) on total area cultivated, maize area, and
non-maize cultivated area. Results show that cultivated area declined in the short-term
by 0.32 hectares (ha) among all households afflicted by PAM, but no medium or long-
terrn effects were significant. Non-maize area declined in the short-term by 0.23 ha while
maize area declined in the medium term by 0.33ha among households afflicted by PA
death. Reduction in planted area was greatest after a male household head or spouse died.

The decline in staple maize area in the medium term has negative implications for

household food security. We recommend provision of input credit (for fertilizer and
hired labor) for maintaining non-maize crops in the short-term. Afflicted households
could mitigate food insecurity with cash proceeds from marketing non—maize crops.

The second essay uses a double hurdle model composed of a first-stage probit
model followed by a truncated regression on non-zero observations from 2002 cross-
sectional survey data in order to assess PAM impact on participation and intensity of
participation in rural non-farm employment. Results show that PAM affects the
decisions of surviving adults on participation in rural non-farm employment, depending
on the gender and household position of the deceased. A “labor pull” effect occurs when
a PA female head or spouse dies, as home care labor demand causes retraction of labor
from non-farm to farm household activities. A “financial push” effect occurs when the
death of a PA male occurs, as surviving adults enter the non-farm employment market to
recover finances lost during illness and funerals. Death of PA women reduced likelihood
of participation in agricultural rural non-farm employment (RNFE) and net in the non—
agricultural RNFE, suggesting potential gender barriers to entry into high pay—off non-
agricultural markets. Surviving adults are more likely to participate in RNFE in the short
term (0-6 years) after a PA death and less likely over the long term (7-13 years). For
those who participate in the non-farm employment, intensity levels drop in the days
following the death shock but increase over time. Injecting credit for labdr hire and
inputs in the short-term could increase productivity of retracted labor in the short term
and help prevent the observed medium term decline in maize area planted. For severely
afflicted households, this intervention strategy needs to be complemented with direct

food aid, as part of a comprehensive strategy to strengthen local social safety nets.

Copyright by
EDWARD PEPUKAYI MAZHANGARA
2007

DEDICATION

To Nyasha and Fadzai, the best cheerleaders in my squad.
To Tendi and my late father Tazvisiya Weston, who both wanted me to be a doctor!
To my mom, Suzanne, a precious gift I cherish always!
To Dr. Emmanuel Manzungu, a great mentor and life coach!
To Jane and Washington Mutatu, love and kindness in action.
To my brother Don Slocum and my Slocum family in Ohio, a pillar of strength!
And most of all to My Lord, Savior and Master, Jesus Christ, for carrying me through
many storms, my answer is Yes Lord!

ACKNOWLEDGEMENTS

I am thankful for the tuition and stipend support I receive from The Rockefeller
Foundation and grants and assistantships I received from the Department of Agricultural
Economics throughout my tenure at Michigan State University. I thankfully acknowledge
financial support from an MSU/Compton Foundation Peace Fellowship, the MSU Food
Security II project, and the USAID Malawi mission for the dissertation fieldwork.

I am greatly indebted to my guidance committee members: Scott Swinton, Eric
Crawford, John Goddeeris, and Cynthia Donovan. Earlier phases of my work benefited
immensely from contributions by John Strauss before he transferred to the University of
Southern California.

I would like to thank the wonderful farmers I met, interviewed and interacted with
in Mzuzu, Lilongwe and Blantyre Agricultural Development Divisions, in Malawi. Their
wealth of knowledge, patience and hospitality was unparalleled. I now know why Malawi
is called the “Warm heart of Africa”.

I would also like to thank the following people who made it possible for me to
conduct research in Malawi: Melinda Smale, Larry Rubey, Dickxie Kampani, Mike
Weber, Dave Wiley, Anne Ferguson, Todd Benson, Kimberly Smiddy, the late Wyclife
Chilowa and staff at Center for Social Research in Zomba, Charles Machinjili, Charles
Mataya and the staff of Agricultural Policy Research Unit — Bunda College of
Agriculture University of Malawi, Stanley Khaila, Hardwick Tchale, Misheck Mtaya,
Francis Sauli, P.A.Z. Mtegha, M.N.S. Msowoya, the late Hendrick Sagawa and my field

enumerator staff.

vi

While at MSU in the graduate school, I was privileged to interact with fellow
graduate students. You all left an indelible mark on my life. The friendships made in G5
will last a lifetime. Thank you for your support and encouragement.

I am grateful to Pastor. Dave Williams — President of Mount Hope Bible Training
Institute and his staff; Mark Bender, Academic Dean, Tim Backus, Dean of Students, and
Lisa White, Registrar. My life was greatly enriched by instructors and students, during
my 2-year hiatus (2004—2005) from MSU while pursuing a Diploma in Ministerial
Studies.

Outside of campus life I have had the joy of meeting and interacting with
wonderful people who have enriched my life. I am grateful to have known the following
Steve Tass, Dan Camcross, Steve Vanis, Jodie and Jessica, Carl Schmidt, Dave Swick,
Roger Schneeberger, Richard Snover, Jeff Bamas, Yvonne Takyi, Al Sharp and Bishop
Anthony Yeboah, to mention but a few.

Last but not least, I want to acknowledge, the support of a very unique individual,
a friend and a mentor who poured so much in my life in so many ways, too numerous to
mention. He played so many roles in my academic stay while I was at MSU. There were
times I had so much to say but did not have to because he seemed to understand it all. I
could not have done it with his supportive role and guidance. That is none other than my
Advisor, Professor Scott M. Swinton. From the bottom of my heart thanks a Million

Prof!

vii

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. xi
LIST OF FIGURES ........................................................................................................... xv
LIST OF ACRONYMS .................................................................................................... xvi
CHAPTER 1
INTRODUCTION ............................................................................................................... l
1. 1 Background .................................................................................................... l
1.2 Problem Statement ......................................................................................... 2
1.3 Research Objectives and Questions ............................................................... 3
1.4 Organization of the Dissertation .................................................................... 5
References ........................................................................................................................... 7
CHAPTER 2
DATA SOURCES AND FIELD RESEARCH METHODOLOGY ................................... 9
2. 1 Introduction .................................................................................................... 9
2.2 Study Area ...................................................................................................... 9
2.3 Data Sources ................................................................................................. 13
2.4 Sampling ...................................................................................................... 15
2.4.1 Focus Group Sampling ................................................................................. 15
2.4.2 Formal Survey Sampling ............................................................................. 17
References ......................................................................................................................... 20
CHAPTER 3
IMPACT OF ADULT MORTALITY ON AREA PLANTED BY RURAL FARM
HOUSEHOLDS IN MALAWI ......................................................................................... 21
3.0 Introduction .................................................................................................. 2 l
3.1 HIV/AIDS Pandemic in Southern Africa and Malawi ................................. 23
3.1.1 Overview of Malawian Agriculture and HIV/AIDS .................................... 23
3.2 Review of Literature .................................................................................... 25
3.2.1 Early Studies on HIV/AIDS and related impacts ......................................... 26
3.2.2 Recent Empirical Studies of the Effect of PA death on Agricultural
Production .................................................................................................... 27
3.3 Theoretical Model ........................................................................................ 30
3.4 Data and Model ............................................................................................ 37
3.4.1 Data .............................................................................................................. 37
3.4.2 Sampling Method and Procedure ................................................................. 38
3.4.3 Sample Size, Attrition and Comparison of Means ....................................... 39
3.5 Estimation Strategies and Empirical Model ................................................. 43
3.5.1 Estimation Strategies .................................................................................... 43
3.5.2 Econometric Model ...................................................................................... 48
3.5.3 Variable Construction .................................................................................. 51
3.5.4 Econometric Concerns ................................................................................. 57

viii

3.6 Results .......................................................................................................... 61
3.6.1 Characteristics of Affected PA Individuals and Households ....................... 62
3.6.2 Results for Endogeneity Test ....................................................................... 68
3.6.3 Impact of PA Mortality on Total Cultivated Area in Malawi ...................... 68
3.6.4 Impact of PA Adult Mortality on Maize and Non-Maize Area Planted ...... 72
3.6.5 The Role of Knowledge and Labor Losses in Changes in Area Planted

by Rural Malawian Agricultural Households. ............................................. 78
3.6.6 Short term, Medium-term and Long-term Impacts of PA Mortality on

Area Planted among Afflicted Malawian Agricultural Households ............ 82
3.6.7 Effects of PA Mortality on Agricultural Production through Change in

Area Planted among Afflicted Malawian Agricultural Households ............ 93

3.7 Conclusion ................................................................................................... 95

References ....................................................................................................................... 100

CHAPTER 4

NON-FARM LABOR ALLOCATION DECISIONS AMONG MALAWI’S RURAL

SMALLHOLDER HOUSEHOLDS IN AN ERA OF HIV/AIDS .................................. 106

4.0 Introduction ................................................................................................ 106

4.1 Previous Research on Rural Non-Farm Employment and Adult death 108

4.2 Study Objectives ........................................................................................ 113

4.3 Conceptual Model ...................................................................................... 114
4.3.1 Modeling Non-farm Work Decisions ......................................................... 121
4.3.2 Modeling Joint Versus Separate Participation and Intensity Decisions ..... 123

4.4 Hypotheses ................................................................................................. 128
4.4.1 Decision Making Process Hypothesis ........................................................ 129
4.4.2 Participation Hypotheses ............................................................................ 129
4.4.3 Level (Intensity) of Participation Hypotheses ........................ . ................... 131

4.5 Empirical Methods and Data ...................................................................... 132
4.5.1 Data and Econometric Model ..................................................................... 132
4.5.2 Empirical Model for participation in Aggregate RNFE, Agricultural

and Non-agricultural RNFE ....................................................................... 134
4.5.3 Empirical Model for Level of Intensity of Participation in Agriculture

and Non-agricultural RNFE ....................................................................... 144
4.5.4 Econometric Concerns ............................................................................... 144

4.6 Regression Results ..................................................................................... 145
4.6.1 Results on Univariate Analysis of Types of Non-farm Activities ............. 145
4.6.2 Results on Tobit vs Double Hurdle Model ................................................ 160
4.6.3 Determinants of Individual Participation in RNFE .................................... 162
4.6.4 Determinants of Intensity Levels of Individual Participation in RNFE ..... 172
4.6.5 Summary Discussion Comparing the Role of Death Variables in

Influencing Decisions on Participation and the Intensity Levels in

RNFE Markets. .......................................................................................... 180
4.7 Conclusions ................................................................................................ 181
References ....................................................................................................................... 1 85

ix

CHAPTER 5

SUMMARY OF FINDINGS AND POLICY IMPLICATIONS .................................... 189
5. 1 Background ................................................................................................ 189
5.2 Summary of Findings ................................................................................. 191
5.2.1 The Impact of Prime age Mortality on Agricultural Production ................ 192
5.2.2 The Impact of Prime-age Mortality on Rural Non-farm Employment ...... 194
5.3 Policy Implications and Future Research Needs ........................................ 195
References ....................................................................................................................... 199
APPENDICES ................................................................................................................. 200
Appendix Al: Focus Group Discussion Checklist ..................................................... 202
Appendix A2: MAPAMS Survey Instruments .......................................................... 206

Appendix A3: A Dialogue on HIV/AIDS, Adult Death and Agriculture with Rural
Communities in Malawi: Highlights from Field Notes Collected during

Focus Group Discussions ........................................................................... 260
Appendix B: HIV/ADS Pathways of Influencing Rural Households ......................... 302
Appendix C: Model Selection — Double Hurdle vs Tobit and Biprobit vs Probit ....... 304
Appendix D: Death Timing Model Results: rpam, epam and YSD ............................ 309

Table 2.1

Table 2.2

Table 3.1

Table 3.2

Table 3.3

Table 3.4

Table 3.5

Table 3.6

Table 3.7

Table 3.8

Table 3.9

Table 3.10

Table 3.11

LIST OF TABLES

The study area for Malawi Agricultural Production and Adult
Mortality Survey (2002): Region, Agricultural Development
Division and Enumeration Area .................................................. 12

Number of households by region in rural Malawi between
1990 and 2002 ...................................................................... 18

Number of households, prevalence of prime-age morality, by
province in rural Malawi between 1990 and 2002 ............................ 40

Initial (1990) household characteristics stratified by attrition
status ................................................................................. 42

Pre-incidence (1990) household characteristics in Malawi
stratified by PA mortality status .................................................. 45

Post-incidence (1990) household characteristics in Malawi
stratified by PA mortality status .................................................. 46

Illustration of the difference-in-indifference Estimation Approach ........ 47

Characteristics of deceased and non-deceased prime-age
individual in survey area of Malawi, 1990-2002 ............................... 63

Ex-ante (1990) and ex-post (2002) household characteristics of
afflicted and non-afflicted households in survey area of
Malawi .............................................................................. 65

Statistical differences between PA death households and NPA
death households ................................................................... 67

The impact of PA mortality, gender and household membership
position of the deceased on total area cultivated in Malawi (ha),
results of difference—in-difference regression, 1990 —2002 .................. 70

The impact of PA mortality, gender and household membership

position of the deceased on maize and non-maize cultivated

areas (ha), results of OLS difference-in-difference regression,

1990 —2002 ........................................................................ 73

The impact of PA mortality by gender and headship or

spousal position in the household on maize and non-maize

cultivated areas (ha), results of OLS difference-in-difference

regression, 1990 —2002 .......................................................... 76

xi

Table 3.12

Table 3:13

Table 3.14

Table 3.15

Table 4.1

Table 4.2

Table 4.3

Table 4.4

Table 4.5

Table 4.6

Table 4.7

Table 4.8

The impact of PA mortality loss of labor, knowledge and

money of total area cultivated, maize and non-maize area

cultivated in Malawi, results of OLS difference-in-difference

regression, 1990 —2002 .......................................................... 79

The Impact of PA mortality by position in household by
gender and period of death on total Area planted......... . . 83

The Impact of PA mortality by position, gender and period
of death on maize and non-maize area planted ................................. 88

Average maize yields in kilograms per hectare estimated from
yield sub plots in Blantyre, Kasungu and Mzuzu ADDS .................... 94

Classification of types of rural non-farm employment activities
from Rural Malawi ................................................................ 112

Descriptive statistics of variables using the Double Hurdle
and Tobit regression analysis of survey data from rural
Malawi ............................................................................ 135

Adult individual participation in types of non-farm employment
Activities available in rural Malawi, 2002 .................................... 146

Adult individual participation in types of non-farm employment
activities by gender in rural Malawi, 2002 .................................... 147

Adult individual participation in main non-farm employment
activities by region in rural Malawi, 2002 .................................... 149

Comparative statistics of participants and non-participant
individual’s data used in empirical estimations of rural non
farm employment participation (RNFEp) in Malawi, 2002 .............. 153

Comparative statistics of participants in AgRNFE and

Non-agRNFE for individual’s data used in empirical estimations

of rural non farm employment participation (RNFEp) in

Malawi, 2002 ..................................................................... 157

Determinants of participation in rural non-farm employment

by adult individuals in rural households in Malawi, 2002

MAPAMS study. Probit Model Estimation (with PA

Death Variables) ................................................................. 163

xii

Table 4.9

Table 4.10

Table 4.11

Table A3.]

Table A3.2

Table A3.3

Table A3.4

Table A3.5

Table A3.6

Table A3.7

Table A3.8

Table A3.9

Table A3.10

Determinants of Participation in rural non—farm employment

by adult individuals in rural households considering timing of

PA death in Malawi, 2002 MAPAMS study. Probit Model

(with Timing of PA Death Variables) ......................................... 166

Determinants of Intensity of Participation in rural non-farm

Employment by adult individuals in rural households considering

timing of PA death in Malawi, 2002 MAPAMS study. Truncated
Regression (with PA Death Variables) ....................................... 173

Determinants of Intensity of Participation in rural non-farm

Employment by adult individuals in rural households considering

timing of PA death in Malawi, 2002 MAPAMS study. Truncated
Regression (with Timing of PA Death Variables) ........................... 176

Characteristics of the wealth ranking categories in Thumbwe
And Kamwendo numeration Areas, in Southern Region

of Malawi ........................................................................ 269

A Ranking of main sources of livelihood Activities in the
Thumbwe and Kamwendo Enumeration Areas (EAs) ..................... 271

Lists of constraints to farming in Thumbwe and Kamwendo
Enumeration Areas .............................................................. 272

A ranking of the main illnesses that cause death in Thumbwe
and Kamwendo Enumeration Areas ........................... ' ................ 273

Things considered key assets in Thumbwe and Kamwendo
Enumeration Areas .............................................................. 274

Characteristics of the wealth ranking categories in Kamwendo
and Kavuta Enumeration Areas, in Central Region of Malawi ............ 280

A ranking of main sources of livelihood activities in the
Kamwendo and Kaluluma Enumeration Areas (EAs) ..................... 281

Lists of constraints to farming in Kamwendo and Kaluluma
Enumeration Areas .............................................................. 282

A ranking of the main illnesses that cause death in Kamwendo
and Kaluluma Enumeration Areas ............................................. 283

Things considered key assets in Kamwendo and Kaluluma
Enumeration Areas ............................................................. 284

xiii

Table A3.11

Table A3. 12

Table A3.13

Table A3.14

Table A3.15

Characteristics of the wealth ranking categories in Kamchocho
and Mzalangwe EAs, in Northern Region of Malawi ....................... 289

A Ranking of main sources of livelihood activities in the
Kamchocho and Mzalangwe Enumeration Areas (EAs) ................... 291

List of constraints to farming in Kamchocho and Mzalangwe
Enumeration Areas ............................................................... 292

A ranking of the main illnesses that cause death in Kamchocho
and Mzalangwe Enumeration Areas .......................................... 293

Things considered key assets in Kamchocho and Mzalangwe
Enumeration Ereas ................................................................ 293

xiv

Figure 2.1

Figure 3.1

Figure 4.1

Figure Bl.

LIST OF FIGURES
Map of Malawi showing study area locations of Blantyre,
Kasungu and Mzuzu ADDS ...................................................... 11

PA mortality variables indicating disaggregating by gender,
position in household ............................................................. 53

Mean annual agRNFE and non_agRNFE per capita income by
gender and ADDs ............................................................... 151

Potential pathways by which HIV /AIDS prime-age morbidity and
mortality affects rural farm households ..................................... 303

XV

ADD
ADMARC
AIDS
APRU
CARE
CIMMYT
EA

FGD
GVH

HIV
IFPRI
MAI
MAPAMS
MOA

MP
MVTS
NACP
NGO

PA

PAD
PAM
PRA

RDP
RNF
RNFE
TIP
UNAIDS
USAID
WHO

LIST OF ACRONYMS

Agricultural Development Division

Agricultural Development and Market Cooperation
Acquired Immune Deficiency Syndrome
Agricultural Policy Research Unit

Co-operative Relief and Assistance Everywhere
lntemational Maize and Wheat Improvement Center
Enumeration Area

Focus Group Discussion

Group Village Headman

Human Immune Virus

lntemational Food Policy Research Institute
Ministry of Agriculture and Irrigation

Malawi Agricultural Production and Adult Mortality Study
Malawi Ministry of Agriculture

Member of Parliament

Maize Variety and Technology Adoption Survey, 1989-91
National AIDS Control Programme
Non-Govemmental Organization

Prime-age

Prime-age Adult Death

Prime-age Adult Mortality

Participatory Rural Appraisal

Rural Development Program

Rural Non-Farm

Rural Non-Farm Employment

Targeted Input Programme

Joint United Nations Programme on HIV/AIDS
United State Agency for lntemational Development
World Health Organization

xvi

CHAPTER 1

INTRODUCTION

1.1 Background

Mortality of sexually active and highly productive prime-age adults has increased
significantly with the advent and spread of the Human Immunodeficiency Virus (HIV),
the virus that causes acquired immune deficiency syndrome (AIDS). Originally viewed
as a health sector problem, AIDS has become a pandemic that is recognized as a multi-
sectoral development challenge around the world and in particular Sub-Saharan Africa.
UNAIDS (2006) estimates that 38.6 million people are living with HIV including 25.8
million in Sub-Saharan Africa at the end of 2005. In 2005 alone, 2.8 million are
estimated to have died from HIV/AIDS related illness. In Southern Africa, more than
27% of the adult population is currently living with AIDS. In Malawi, a largely agrarian
economy, the national prevalence rate is 15 percent.

Since the advent of the pandemic, researchers have grappled to understand the
impacts of the resulting adult morbidity and mortality, with a view to developing
measures to mitigate the negative effects of prime-age adult death. Earlier studies
conducted in the 1980’s and 1990’s based on macro-modeling (Cuddington, 1993a,b),
cross sectional surveys and case studies indicated that AIDS could adversely affect labor
supply and remittances to agriculture, causing a possible shift from labor-and capital-
intensive cash crops to low-labor-intensive food and subsistence crops (Barnett, et. a1,
1995). The initial round of anecdotal evidence and earlier formal surveys generally

lacked counterfactual analysis (Mather, et a1, 2005). However, these gave way to more

rigorous focused studies using quantitative methods to capture inter-temporal effects of
the long-term nature of AIDS from infection to death. The Kagera, Tanzania, AIDS
impact studies used panel data methodology and showed that at the household level an
adult death does not necessarily equate to loss of labor, as other family members often
joined the household to take responsibility (Ainsworth and Semali 1998, Beegle 2003).
Yamano and Jayne (2004) and Chapoto (2006) found that significant changes in cropping
patterns occurred for households in the lower half of the income distribution with clear
effects by gender and position of the deceased in the household.

Earlier studies made recommendations for intervention, the most prominent one
being advocating labor-saving technologies. More recent studies have recommended
diversifying sources of livelihood rather than labor replacement. Jayne et a1. (2005)
caution that labor saving interventions may be appropriate for certain types of households

and regions but not across the board.

1.2 Problem Statement

A review of the literature on adult death in the era of AIDS reveals that, while
AIDS has been widely characterized as a large-scale, long-wave, cross—cutting pandemic,
there have been few (but increasing) studies that try to estimate long-term impacts
(Beegle, et a1 2006, Drinkwater, et. a1, 2006, Chapoto, et. al, 2006). The majority of
studies tend to focus on short to medium term (4-6 year) periods due to data scarcity.
The long—term effects of prime age adult death have been little examined and are not well
understood. Three major research questions arise from this gap. First, what is the impact

on agricultural production of prime-age adult mortality (PAM) induced losses of human

and financial capital? Second, what is the impact of PAM on afflicted households’ labor
participation in the rural non-farm sector? From these questions follows a third one
related to intervention strategies to ameliorate the implication of PAM on rural
agricultural households. Haddad and Gillespie (2001) posed this third question as
follows, “How should government policies in the area of food security and agriculture be
better altered to meet the needs of the poor within the context of the HIV/AIDS
pandemic?”

This dissertation attempts to fill these research gaps by providing empirical
evidence of the impact of prime-age adult death on rural agricultural households.
Focusing on both the farm and non-farm sectors, it uses a 13-year panel data set from

Malawi as well as a recent cross-sectional data set.

1.3 Research Objectives and Questions

The broad goal of this study is to assess the impact of prime age adult death on
Malawi’s rural agricultural households in order to inform a strategic response on
mitigation measures.

The study has three specific objectives and associated research questions:
1. To assess the impact of prime age adult mortality on area cultivated and
agricultural production and relative effects from loss of human and financial
capital. The associated research questions are:

a) Does PA mortality affect household-level area cultivated?

b) Do the gender and position of the deceased PA individual make a
difference in terms of magnitude and direction of death impact on area
cultivated?

c) Do PA death-induced losses of financial and human capital impact on area
cultivated? If so, can the magnitudes and direction of impact help us
formulate empirically based interventions to mitigate the impacts?

d) How is the area cultivated and agricultural production of households
afflicted by PA mortality affected by the death shock in the short-term,

medium—term and long term?

. To evaluate how death stricken households respond to rural non-farm
employment (RNFE) labor markets and how their response changes over an

adjustment period since the initial death shock. The associated research questions

are:

a) What are the determinants of an individual’s participation in agricultural
and non-agricultural RN FE in rural Malawi?

b) What are the determinants of the level of intensity in participation

measured by level of labor allocation in agricultural and non-agricultural
RNFE?
c) Is there a difference between initial response and response to death after some
period? How long is the adjustment period, and what can we learn about

household resiliency over time?

3. To recommend policy responses that ameliorate the negative impacts of PA
mortality (PAM) on rural households in Malawi as follows:

a) Given empirical evidence on the impact of PAM on area cultivated and
agricultural production in rural Malawi, what policy measures (if any) need to
be formulated to support agricultural production in the era of HIV/AIDS?

b) Given empirical evidence of determinants of participation and intensity, what
policy measures (if any) need to be formulated to support participation in

RNFE in the era of HIV/AIDS?

1.4 Organization of the Dissertation

This dissertation is organized into five chapters, including this introduction.
Chapter Two provides a description of study area, data sources and field research
methods used, as well as survey sampling methods and the contents of the survey
instruments.

Chapter Three is a self-contained essay that addresses the first objective and first
part of objective three. It begins with an overview of AIDS and Malawian agriculture and
a review of the AIDS impact literature. A section containing a theoretical model and
empirical methods follows this part. We use difference-in—differences econometric
estimation to assess the impact of PA adult death on area cultivated and agricultural
production, measured by total area planted, and area planted to maize and non-maize
crops in Malawi. Finally the chapter addresses impacts of human and financial capital
losses on area planted and provides specific policy recommendations to mitigate negative

impacts.

Chapter Four is a second self—contained essay that addresses the second research
objective and the second part of objective three. The essay opens with highlights of
research on PA death and rural non-farm employment, plus a conceptual framework of
analysis. We then use a double hurdle econometric model to estimate the determinants of
individual participation and intensity levels in aggregate RNFE, agricultural RNFE and
non-agricultural RNFE. Finally the chapter discusses the role of death variables in non-
farm labor allocation decision among rural households on Malawi and assesses the
related policy implications.

Chapter Five summarizes the conclusions and economic policy implications

before making recommendations for further research.

References

Ainsworth, M., and I. Semali. (1998). “Who is most likely to die of AIDS? Socio-
economic Correlates of Adult Deaths in Kagera Region, Tanzania”. In
Confronting AIDS: Evidence from the Developing World, (ed). M. Ainsworth, L.
Fransen, and M. Over 95-110. Washington, DC: World Bank.

Barnett, T., J. Tumushabe, G. Bantebya, R. Sebuliba, J. Ngasongwa, D. Kapinda, M.
Ndileke, M. Drinkwater, G. Mitti, M. Haslwimmer (1995). “Final Report: The
Social and economic Impact of HIV/AIDS on farming systems and livelihoods in
rural Africa: Some experience and lessons from Uganda, Tanzania and Zambia.”
Journal of lntemational Development 7, 163-176.

Beegle, K, (2003), “Labor Effects of Adult Mortality in Tanzanian Households”, World
Bank Policy Research Working Paper no. 3062, World Bank, Washington, DC.

Beegle, K., J. De Weerdt, and S. Dercon. (2006). “Adult Mortality and Economic Growth
in the Age of HIV/AIDS. Draft paper. World Bank, Washington, DC.

Chapoto, A. (2006). “The Impact of AIDS-related Prime-Age Mortality on Rural Farm
Households: Panel Survey Evidence from Zambia”. PhD Dissertation.

Department of Agricultural Economics, Michigan State University, East Lansing,
MI 48824. USA

Chapoto, A. and T. Jayne (2006). “Impact of AIDS-Related Mortality on Rural Farm

Households in Zambia: Implications for Poverty Reduction Strategies.” Mimeo.
Michigan State University, East Lansing, MI. USA.

Cuddington, J ., (1993a) “Modeling the macroeconomic effects of AIDS, with an
application to Tanzania.” World Bank Economic Review 7, 173 — 189.

Cuddington, J ., ( 1993b) “Further results on macroeconomic effects of AIDS: The
dualistic labor surplus economy.” World Bank Economic Review 7, 403 — 417.

Drinkwater, M., M. McEwan, and F. Samuels. (2006). ”The Effect of HIV/AIDS on
Agricultural Production Systems in Zambia: A Restudy 1992-2005. Analytical
Report. www.ifpri.org/renewal/pdf/Zambia_AR.pdt

Haddard, C. and S. Gillespie. (2001). “Effective Food and Nutrition Policy Response to
HIV /AIDS. What we know and what we need to know.” Journal of lntemational
Development 13, 487—511.

Jayne, T. S., M. Villarreal, P. Pingali and G. Hemrich. (2005). “HIV/AIDS and the
agricultural sector in Eastern and Southern Africa: Anticipated Consequences.” A
Paper presented at the lntemational Conference on HIV/AIDS and Food and
Nutrition Security, Durban, 14—16 April 2005.

Mather, D., C. Donovan, T. Jayne and M. Weber. (2005). “Using Empirical Information
in the Era of HIV/AIDS to Inform Mitigation and Rural Development Strategies:
Selected Results from African Country Studies”. American Journal of
Agricultural Economics 87 (5): 1289-1297.

UNAIDS. “Report on the Global HIV/AIDS Epidemic 2006”. UNAIDS, Geneva,
Switzerland. http://wwwunaidsorg/en/HIV%5Fdata/2006GlobalReport/

Yamano, T., and T. S. Jayne. (2004). “Measuring the Impacts of Working-Age Adult
Mortality on Small-Scale Farm Households in Kenya.” World Development, 32
( 1): 91-1 19.

CHAPTER 2

DATA SOURCES AND FIELD RESEARCH METHODOLOGY

2.1 Introduction

This chapter presents the study area, data sources and some field research
methodological issues pertaining to this study. The data come from two studies, a
baseline Maize Variety and Technology Adoption Survey (MVTS) conducted by
lntemational Maize and Wheat Improvement Center and Malawi Ministry of Agriculture
(CIMMYT/MOA) in 1989/90 and a follow up Malawi Agricultural Production and Adult
Mortality Study (MAPAMS) conducted in 2001/2002. MAPAMS restudied 351
households from the MVTS to construct a l3-year longitudinal panel data set. That data
set enables control for household fixed effects in analyzing death impacts on farm
production in Chapter Three. Chapter Four uses only the cross-sectional data set from
the 2001/2002 MAPAMS survey to analyze death impacts on non-farm labor allocation

decisions. Below are details on the study area, data sources and survey sampling.

2.2 Study Area

Three criteria guided the selection of the study area for this dissertation; (i) a
country in Southern Africa region where the HIV/AIDS prevalence rate was high and
hence adult mortality expected to be high, (ii) genuine interest by policy makers to gain
insight into prime age adult mortality impacts on agriculture to effect suitable policy
change and (iii), availability of good baseline data to enable a restudy and construct a

panel data set. A review of Joint United Nations Programme on HIV/AIDS (UNAIDS)

fact sheets (UNAIDS, 1999, UNAIDS/WHO, 2000) and literature from Malawi (Bisika
and Kakhongwe, 1995; NACP, 1999-2001), consultations with policy makers during a
pre—dissertation visit to Malawi and the availability and granting of permission and access
to a good benchmark dataset by CIMMYT/MOA, the MVTS,I(Smale, eta], 1991)
resulted in selection of Malawi as the study area.

The study was conducted in three regions of Malawi, namely, Southern, Central
and Northern Regions. The study covered one Agricultural Development Division
(ADD) in each region, representing the main maize producing areas of Malawi: Blantyre
ADD in the Southern Region, Kasungu ADD in the Central and Mzuzu ADD in the

Northern Region shown in the map in Figure 2.1 below.

10

 

 

 

 

L— #__ if; ,2 , ,,.;__—____ . , ,_. ._ 7.

Figure 2.1 Map of Malawi showing study area locations of Blantyre, Kasungu
and Mzuzu ADDS

The seven enumeration areas (EAs) in each ADD that were visited in the MVTS

baseline survey were revisited under MAPAMS. Attempts were made to locate the

original 20 households per EA. Table 2.1 displays the study areas by region, ADD and

enumeration area.

 

 

 

 

 

Table 2.1 The study area for Malawi Agricultural Production and Adult
Mortality Survey (2002): Region, Agricultural Development Division
and Enumeration Area.

Region Agricultural Enumeration Areas

Development Division (EAs)
(ADD)
Southern Blanytre ADD Neno, Mwanza RDP*
Ngadziwe, Mwanza RDP
Chanasa, Phalombe RDP
Kamwendo, Mulanje RDP
Thumbwe, Chiradzulu RDP
Mbulumbuzi, Chiradzulu RDP
Bvumbe, Thyolo RDP
Central Kasungu ADD Kaluluma, Kasungu North RDP
Lukwa, Kasungu RDP
Chakhaza, Dowa West RDP .
Dzoole, Dowa West RDP
Msakambewa, Dowa West RDP
Kabvuta, Mchinji RDP
Zulu, Mchinji RDP
Northern Mzuzu ADD Mpherembe, Northern Mzimba RDP

 

 

Kapando, Central Mzimba RDP
Manyamura, Central Mzimba RDP
Kafukule, Central Mzimba RDP
Mzalangwe, Central Mzimba RDP
Emcisini, North Mzimba, RDP
Kamchocho, Central Mzimba, RDP

 

Source: .MVTS 1990, MAPAMS 2002

Notes

* RDP stands for Rural Development Program. This is an administrative boundary
whereas an EA is a Central Statistical Office construct.

12

 

2.3 Data Sources

The first Malawi AIDS case was recorded in 1985, placing the pioneer AIDS
mortality victim around 1992. In order to obtain a longer term panel data set that could
possibly capture AIDS effect, it was determined we find a benchmark data set from a
survey conducted prior to 1992, which contained good demographic and agricultural
production data for revisiting to capture demographic, agricultural and mortality data.

MVTS suited that purpose well.

MVTS data (1990)

The objectives of the CIMMYT/MOA MVTS were to profile farmer adoption of
maize varieties and to assess how household factors influenced farmers’ selection of
maize varieties (Smale, et a1 1991). It was designed as a module attached to Malawi’s
Annual Survey of Agriculture (ASA) and it was conducted on a subset of 420 households
included in the 1989-90 national sampling frame. The survey collected data on
agronomic practices, maize varietal information, production costs, yield levels, wages,
prices, rainfall records and household income (Smale, et a1 1993). Re-establishing
contact with households was feasible because of institutional linkages between
lntemational Maize and Wheat Improvement Center (CIMMYT) and Malawi’s Ministry
of Agriculture (MOA) and the good record keeping of the institutions, researchers and

enumerators involved in the survey.

13

Recent data (2002)

The follow-up fieldwork in 2002 began with focus group discussion, a qualitative
method, to understand the context in which prime age adult death was understood. Those
discussions were used to develop the survey instrument for the MAPAMS interviews of
the MVTS respondent panel. Both the qualitative and quantitative methods are discussed

below.

Focus Group Discussion

The informal survey took the form of focus group discussions (FGD) designed (i)
to obtain a wider picture of the current state of agricultural productivity, adult mortality
and rural livelihoods in each region from more than just the MTVS households, (ii) to
obtain understanding and clarification on conflicting views on issues that emerged from
formal questionnaires, (iii) to involve the local traditional leaders other key informants,
(iv) to bring together different age and gender groups in order to assess Views on
divergent issues, and (v) to provide a forum for picking up pertinent issues that would
otherwise not be picked up by the formal survey questionnaires. The checklist for the

focus group discussions is available in Appendix A1.

MAPAMS

The Malawi Agricultural Production and Adult Mortality Study formal survey
obtained information on demographics, farm household assets, management practices,
yield and area measurements and death history collected using a set of five instruments.

The first instrument, MAPAMS l — Roster book, collected information on household and

14

demographic characteristics including death history in the household. The second
instrument, MAPAMS 2 — Plot Book, collected information on agronomic practices from
land preparation to harvesting and yield measurements. Agronomic data was collected in
two visits, after second weeding and at the end of the harvest.lThe next instrument,
MAPAMS 2S — Plot Book Garden Measures, is a supplementary component of the plot
book that captured cropped field area measurements. The fourth instrument, MAPAMS 3
— Livelihoods Book, collected data on assets, non-farm employment and livestock. The
last instrument, MAPAMS 4 — GPS Entry Book, collected information on GPS

coordinates. The MAPAMS instruments are attached in Appendix A2.

2.4 Sampling

2.4.1 Focus Group Sampling

Two enumeration areas (EAs) in each ADD were purposefully selected based on
the characteristics of highest number of households affected by adult mortality. For each
of these EAs, all “MVTS” farmers were invited. Further 20 — 25 other non-project
farmers were randomly selected from the same villages to make up a total of 40
participants for each FGD. The village headmen from those villages were invited to
participate in their capacity as village headmen along with the Group Village Headman'.
Extension personnel from the area were also invited to help with addressing concerns that
relate to their department. Also present were scribes or recorders who understood the
local languages of Chichewa or Chitumbuka. They took detailed notes of the

proceedings in English.

 

I Group Village Headman is a headman who presides over 2 to 3 other headmen under him. Villagers can
appeal to GVH if not satisfied with ruling of their local head man (mufumu)

15

Discussions were held in two sessions, a large group session and a small group
session. The large group session was the main session that started right after the
introductions. This was made up of all the participants. The small group session split
participants along gender lines. A woman discussant was engaged to leading the
women’s small group discussions.

As pointed in the introduction, Focus Group Discussions (FGDs) belong to the
informal methods of Participatory Rural Appraisal (PRA) approaches. PRA
methodologies are good at collecting qualitative data and information quickly but tend to
be limited in that the data cannot be analyzed quantitatively. The approaches also tend
not to be representative. For the FGDs, we selected areas of high adult mortality per
enumeration area not close to national/regional boundary areas. However, Kamwendo
Enumeration Area, Mchinji District could be unique because it is traversed by a main
road to the border town of Mchinji and thus has potential for more sexual transactions
with commercial truckers than other EAs. Inferences from FGDs study Could be drawn,
with caution, over main maize producing rural farming areas of each region of Malawi.
A report of the highlights from field notes collected during the FGD is given in Appendix

A3.

16

2.4.2 Formal Survey Sampling

M VTS

The 420 households interviewed for the MVTS were selected using population-
based multistage random sampling from the major maize producing areas of Malawi
which Blanytre, Liwonde, Kasungu, Lilongwe and Mzuzu Agricultural Development
Districts (ADDS). Three ADDS, Blantyre, Kasungu and Mzuzu were purposively chosen
to represent contrasting agro-ecological and household economic zones reflecting factors
hypothesized to affect maize varietal adoption and production. Evaluation officers listed
representative Enumeration Areas (EAS) from each of the ADDS and used systematic
random sampling to select 7 EAs out of 21 in Blantyre ADD, 7 of 10in Mzuzu and 7 of
25 in Kasungu. The survey area therefore represented the higher potential, major maize
producing areas of Malawi (Smale, 1991). In each BA, 20 households were randomly

selected.

MAPAMS

For the MAPAMS survey in 2002, households were revisited, and Table 2.2
shows the number of households that were located after 13 years. We were able to
interview 351 of the original 420 households interviewed in 1990, for an attrition rate of
16 percent. The greatest attrition was in Blantyre ADD (24 percent rate). The central and

northern regions each had 13 percent attrition.

l7

Table 2.2 Number of households by region in rural Malawi between 1990 and

 

 

 

 

 

 

 

2002
Region Representative Number of Number of Percentage of
ADD Households Households re- missing or
interviewed in interviewed in dropped
1990 2002 (attrited)
households
(3) (b) (C) ((1)
Southern Blantyre 140 107 24
Central Kasungu 140 122 13
Northern Mzuzu 140 122 l 3
Total 420 351 16

 

 

 

 

 

Source: MVTS 1990, MAPAMS 2002

The reasons for the attrition are as follows: Households moved away following
death of a PA adult (2 cases), households moved away for reasons not related to an adult
death such as job search, reuniting with a working spouse, marriage, returning to live
closer to other relatives (43 cases), households dissolved following death of a non PA
adult (8 cases), households dropped for incomplete data sets (10 cases), households with
heads too old to answer questions or suffering from mental disturbance (2 cases),
household not traced (1 case), households dropped out before end of survey in 1990 (2
cases) and a household with death and a married son took over thus constituting a new
household (1 case). The large attrition rate in the Southern Region was largely due to
households that moved to Mozambique to reconnect with their kinsmen after the
Mozambique Civil war had ended.

Analysts worry that attrition may be selective on characteristics like education,
PA death such that a high attrition is likely to bias estimates from longitudinal data

(Alderman eta] 1996). If households that subsequently drop from the sample differ in

18

 

their initial behavioral relationship, from those that do not, then caution should be
exercised when interpreting estimates based only on the nonattriting sample. Potential
bias of the 16 percent attrition for the MAPAMS panel dataset will be discussed in the

next chapter.

19

References

Alderman, H., J .R. Behrman, H. Kohler, J .A. Maluccio and SC. Watkins. “Attrition in
Longitudinal Household Survey Data: Some Tests for Three Developing
Country Samples”. Disscussion Paper 96. Food Consumption and Nutrition
Division, lntemational Food Policy Research Institute, Washington, DC.

Bisika, T.J., and P. Kakhongwe. (1995). “HIV/AIDS, STDs and Skin Diseases in
Malawi: A Review of Literature”. Centre for Social Science Research,
University of Malawi. Zomba. Malawi.

National AIDS Control Programme, ( 1997). “Report on AIDS cases surveillance, 1997”.
Lilongwe: National AIDS Control Programme. Malawi.

National AIDS Control Programme, (1999). “Sentinel surveillance report: HIV/Syphilis
seroprevalence in antenatal clinic attenders”. Lilongwe: National Aids Control
Programme. Malawi.

National AIDS Control Programme, (2000). “Malawi’s response to HIV/AIDS for 2000-
2004”. Lilongwe: Strategic Planning Unit, National AIDS Control Programme.
Malawi.

Smale, M., Z.H.W. Kaunda, H.L Makina, and M.M.M.K. Mkandawire. (1993)
“Farmers’ Evaluation of Newly Released Maize Cultivars in Malawi: A
Comparison of Local Maize, Semi-Flint and Dent Hybrids”. CIMMYT,
Lilongwe, Malawi.

Smale, M., with Z.H.W. Kaunda, H.L. Makina, M.M.M.K Mkandawire, M.N.S.
Msowoya, D.J.E.K. Mwale, and P.W. Heisey. (1991). “Chimanga Cha Makolo,
Hybrids, and CompositeszAn Analysis of Farmer’s Adoption of Maize
Technology in Malawi, 1989-91. CIMMYT Economics Working Paper 91/04.
Mexico, D.F. Mexico.

UNAIDS (1999). “Report on the Global HIV/AIDS Epidemic.” Geneva, June, 1999

UNAIDS/WHO, (2000). Malawi: epidemiological fact sheet on HIV/AIDS and sexually
transmitted infections: Geneva: 2000 update.

20

CHAPTER 3

IMPACT OF ADULT MORTALITY ON AREA PLANTED BY RURAL FARM
HOUSEHOLDS IN MALAWI

3.0 Introduction

Mortality of sexually active and highly productive prime-age (15-59) adults has
increased significantly with the advent and Spread of the Human Immunodeficiency
Virus (HIV), the virus that causes acquired immune deficiency syndrome (AIDS). For
nearly 20 years researchers have been engaged in trying to understand the scope and
magnitude of impact of the HIV/AIDS pandemic and the related adult mortality on
household and national economies especially in developing countries where its spread
has been rampant. The lack of comprehensive and reliable quantitative data hampered
effective empirical analysis. In the recent 5 years more quantitative studies have emerged
to look at the impact of prime-age death on crop production in order to inform policy
responses.

Despite the strides made in recent studies, gaps still exist in the literature with
respect to enumeration of the losses embodied in the passing on of a prime age adult
(PA)2. When a PA individual dies, three productive resources are lost: labor, financial
capital and human capital. The direct labor contribution of the deceased PA individual is
lost and also working days are lost to care for the afflicted and mourn those who die.
Financial capital is lost through money spent on illness and funerals and foregone income
and remittances that combine to become a significant monetary loss whose opportunity

cost may include agricultural production investment. Finally the human capital in the

 

2 Adults whose age range from 15 to 59 years old.

21

form of formal and farming education, marketing contacts, farming experience,
knowledge and skills is lost when the deceased PA was active in farming. Disaggregating
the PA mortality losses can help to prioritize and target policy interventions. A second
gap in the literature is that previous panel data studies tended to cover short time periods
and hence failed to capture recovery periods from the death shocks.

This study contributes to the literature by disaggregating the PA mortality losses
into three categories: labor, financial and human capital. It uses a two-period panel data
set (with 13 years between observations), the longest in the PA death impact literature.
Focusing on rural households in Malawi, the paper addresses the following research
questions:

0 Does PA mortality affect household level area cultivated?

0 Do the gender and position of the deceased PA individual make a difference
in terms of magnitude and direction of death impact on area cultivated?

0 Do PA death-induced losses of financial and human capital impact area
planted? If so, can the magnitudes and direction of impact help us formulate
empirically based interventions to mitigate the impacts?

0 How is area cultivated and agricultural production level of households
afflicted by PA mortality affected by the death shock in the short-term,
medium-terrn and long term?

0 Given empirical evidence on the impact of PAM on area cultivated and
agricultural production in rural Malawi, what policy measures (if any) need to

be formulated to support agricultural production in the era of HIV/AIDS?

22

3.1 HIV/AIDS Pandemic in Southern Africa and Malawi.

UNAIDS (2006) estimates that there were 38.6 million people currently living
with HIV in the world and 25.8 million were in Sub-Saharan Africa at the end of 2005.
In the year 2005 alone, 2.8 million people were estimated to have died from HIV/AIDS
related illness, while the virus newly infected a further 4.1 million. In Southern Africa,
more than 27% of the adult population was currently living with AIDS at the end of
2005. The hardest hit countries are in southern and eastern Africa. These comprise
Botswana, Lesotho, Namibia, South Africa, Swaziland, Zambia and Zimbabwe with an
estimated HIV-prevalence rate exceeding 20% followed by Malawi, Mozambique and
Kenya, with rates between 10 — 20 % (UNAIDS/WHO, 2005).

The 2005 national prevalence rate of 15% in Malawi aggregates and masks the
picture on the ground where the HIV prevalence rates have a narrower yet higher range
of 23.3%-27.9% in urban areas, whilst rural areas have a wider range of 2 —25%, being

highest in the southern region where there is a high population density.

3.1.1 Overview of Malawian Agriculture and HIV/AIDS

Malawi is a rural, developing country in southern Africa that Shares its borders
with Zambia, Mozambique and Tanzania. Its economy is mainly agrarian, with
agriculture contributing 37% of GDP, nearly 90% of employment, 85% of foreign
exchange earnings and 60 percent of the volume of trade. There are three land tenure
systems: a) customary, b) freehold, and c) leasehold. About 6.5 million hectares of

smallholder farms fall under customary tenure system, whereas 1.2 million hectares fall

23

under a combination of freehold and leasehold, in larger estates. Approximately 70-80%
of arable land is under customary tenure, mostly smallholder farming households, 30
percent of which are female headed.

Malawi attained independence from British colonial rule in 1964, and from then
until 1994, Malawi endured 3 decades of one party rule under President Hastings
Kamuzu Banda. Banda maintained the pre-independence dual agricultural sector
enforced by authoritarian rule, trade restrictions and a restriction on smallholder tobacco
production by smallholders. The commercial sector produced tobacco and tea for export;
while the smallholder sector produced maize in surplus to meet his food self-sufficiency
and agricultural surplus export policies. His Slogan, Chimanga Ndi Moyo, maize is life,
saw large investments in extension, credit facilitation, and parastatal market outlets which
resulted in expansion of maize hectarage up to 70% of total area under smallholder
production. Extensification did not increase rural incomes due to depressed, controlled
prices. In 1994, the new government of President Bakili Muluzi lifted tobacco restrictions
and controlled pricing in line with structural adjustment programs of the World Bank.
Smallholder farmers shifted area from maize to tobacco and other crops such as tubers
and legumes. Unfortunately, these changes coincided with the blooming of the
HIV/AIDS pandemic, which grew under the silence of denial by Banda’s regime.

Initially viewed as a health sector problem, HIV/AIDS in Malawi has long ceased
to be seen as purely a health problem but a multi-sectoral problem affecting productivity
and output from all sectors and all walks of life. For Malawi, largely dependent on
agriculture, it becomes imperative to understand the impacts that HIV/AIDS and prime

age death have on that very backbone of the country, because any Shocks to agriculture

24

could have potential negative downstream effects. With 73,000 people estimated to have
died from HIV/AIDS in 2001, while 15% of the adult population was believed to be
living with HIV/AIDS, Malawian policymakers became keen to obtain a clear
understanding of the impacts of prime-age adult death on household agricultural

production in order to develop appropriate mitigatory measures.

3.2 Review of Literature

There has been a decade and half of research on the topic of HIV/AIDS, adult
mortality and the associated economic impacts using a wide range of methodologies.
These have generated a wealth of knowledge. This literature review section is divided
into two sections, the first looks at early studies that cover broadly the topic of HIV/AIDS
and its economic impacts, while the second looks at those recent studies dealing with
adult mortality as it particularly relates to agricultural production.

This essay uses Barnett and Whiteside’s (2002) nomenclature where households
that suffered prime age adult death are labeled as “afflicted” and not “affected”. The
focus is on PA death, because the cause of death is often unknown, and the proximate
cause may be a disease that could have been resisted by a healthy immune system. The

incidence of PA death in southern Africa has sharply increased since the advent of

HIV/AIDS pandemic.

25

3.2.1 Early Studies on HIV/AIDS and related impacts.

Empirical work in the 1980’s and early 19908 focused on macro-level Simulation
modeling to forecast the effects of AIDS on economic growth rates and per capita GDP
under different assumed scenarios of spread of the pandemic (Cuddington 1993,
Cuddington and Hancock 1994, Kambou et.,al. 1992, Over 1992). They predicted a
decline in labor supply that would reduce savings and economic growth. Their work
helped draw attention to a new problem and set the research wheels in motion. In the mid
to late 19905, focus shifted to sectoral (or meso) level impact studies. Studies of the
commercial farm and manufacturing sectors showed that AIDS negatively impacted
output through medical and funeral expenses, lost time due to illness, and the recruitment
and training costs of job tum-over. High costs associated with AIDS were deemed a
threat to long-term viability at the sector level (Bollinger and Stover, 1999, Rugalema, et
al. 1999, Aventin and Huard, 2000; Ndilu et. al, 1998; Fosythe et a1, 1994).

A number of rural household (micro) level studies were launched in Zambia,
Uganda, Rwanda, Tanzania and Kenya parallel to the macro and meso level studies. Data
challenges were enormous and methodology varied from modeling, rapid rural appraisal,
to limited cross-sectional formal surveys that carefully navigated disclosure, stigma and
infringement of privacy issues. The studies produced key insights on the epidemic and its
likely impacts at household level ranging from a decline in land use, crop yields,
livestock production, loss of agricultural knowledge, soil fertility and a rise in
dependency burdens, to shifts towards subsistence cropping (Gillespie 1989a, b, Barnett

1994, Barnett and Blaikie 1992, Barnett et al. 1995, Tibaijuka 1997).

26

The micro level studies were limited in level of rigor and ability to capture inter-
temporal effects of the long-term nature of AIDS from infection to death. The studies
generally used case-study approach, and targeted areas of high HIV/AIDS prevalence.
Some studies lacked a representative non-afflicted population to provide counterfactual
analysis (Mather, et a1, 2005). A more comprehensive empirical study of household level
AIDS impact was the World Bank, Kagera Health Development Survey conducted in
Tanzania, between 1994 and 1996 that focused on the impact of adult mortality on
income, consumption, assets, child nutrition, activities and time allocation, and school
enrollment (Beegle, 1997, Ainsworth and Semali, 1998, and Ainsworth and Dayton,
2003, Ainsworth and Semali, 2000). The Kagera studies introduced panel data
methodology and used illness related adult death in HIV/AIDS prevalent areas as a proxy
for AIDS impacts (Ainsworth, et, al., 1995). The results showed that adult death in a
household does not necessarily equate to loss of adult labor, because another family
member often joined the household to take responsibility. The study implicitly assumed
the opportunity cost of the family member who moved in was zero elsewhere in the

economy.

3.2.2 Recent Empirical Studies of the Effect of PA death on Agricultural

Production

Using panel data collected in Kenya between 1997 and 2000, Yamano and
Jayne, (2004) found that Significant changes in cropping patterns occurred mainly for

households in the lower half of the income distribution with clear effects by gender and

27

position of the deceased in the household. Households with a male head or spouse death
incurred a decline in area cultivated for sugar, tea and horticultural crops, while those
with a female head or spouse death had a significant decline in cereal crop area.

Recent literature on adult mortality impacts has questioned the conventional
wisdom of recommending labor saving technologies for affected households based on a
priori conclusions that they will have lost labor, given empirical studies to the contrary.
Jayne et al (2005) caution that labor saving interventions may be appropriate for certain
types of households and regions but not across the board. Dorward and Mwale (2005)
found labor saving technologies to be harmful if they drive down wage rates already
declining due to HIV/AIDS induced cash constraints on ability to hire. They argued that
direct cash transfers for labor hire maybe the best action that avoids depressing rural
labor markets. In Rwanda, Donovan and Bailey (2005) found high population density
and very small farm sizes, with households using labor replacement strategies versus
labor saving ones by shifting away from erosion controlling crops thus Compromising
long-term soil fertility and sustainable production. They recommended efforts to improve
income generation and diversified sources of livelihood rather than policy shifts to labor
replacement strategies.

While there has been recognition that AIDS is no longer just a health issue but a
large-scale, long-wave, cross-cutting pandemic that creates a need for empirical studies to
understand the impact of PA mortality on the household economy, the long-term
agricultural production and productivity effects of PA mortality have been little
examined. The majority of empirical studies conducted to date have used panel data sets

that cover short to medium term periods (4 — 6 years) that do not capture long-wave

28

implications. A few but growing number of studies are pursuing the use of longer-term
panel data (Drinkwater 2006, Beegle 2006). The Malawi Agricultural Productivity and
Adult Mortality Study (MAPAMS) survey spans over 13 years and therefore makes a
significant contribution to assessing the long-term impacts over and about the short to
medium term ones.

The main objective of this paper is therefore to examine impact of prime age adult
mortality on cultivated area, a production choice variable for afflicted households.
Effects from labor loss, monetary loss and human capital losses will be evaluated
separately. The study follows the example of agricultural literature that used area
cultivated as a proxy for production intent (Houck and Ryan 1972; Morzuch, Weaver and
Helmbeger 1980, Lee and Helmberger, 1985, Chembezi and Womack, 1991). Lee and
Helmberger (1985) demonstrated that expected profit function could be a useful construct
to conceptualize acreage supply response under a regime with a feed grain program.
Using duality theory and Hotelling’s lemma, Lee and Helmberger (1985) Showed that the
corn acreage supply function with respect to corn price maps identically to optimal
expected output differentiated with respect to a constant. Effect of prime age death
would mostly be felt at planting as household decides the area to be planted vs the
resources they have to work the fields. Use of area cultivated, as a proxy for production
is also appropriate given insufficient data on variables such as pests and other weather

variables that could potentially affect actual output.

29

3.3 Theoretical Model

In order to understand the impact of adult death on household welfare measures,
we assume that a rural Malawian household maximizes utility from consumption of a set

of home produced agricultural goods ( X a ), manufactured goods ( X h ), home produced

non-marketable goods ( Xm) and leisure time ( Ll ), conditional on a set of household

exogenous characteristics 2” . We assume that the utility function is concave and
increasing in consumption of a set of the goods and leisure time subject to the prime-age
death status of the household, a labor time constraint, a production function
(technological constraint), and a full income constraint.

The household’s utility maximization model is written as;

, h
Max U(Xa,Xm,Xh,Ll.Z )

xa,xm,xh,1.,
Subject to:

M0: M(D’j's,nDtj’s) (1)
UM0)=Lfa+LW+Lh+Ll, (2)
Q0 =QaiLd/L}a(Mo)+ Li1,vv,A(Mo),Mg(Mo).-zq) A (3)
Pa xa + mem :_ 25" + wLw— NCRD( Mo)+ REMIT( Mo)+Ynh. = Y (4)

Equation (1) is a vector of dummy and continuous variables ( Mo ) that captures

the prime-age adult mortality status of the household identifying whether there has been a

prime-age death (Dtj's =1/0), and identity of deceased prime-age adult individual j by

30

gender ( j = m, f ) and by position or status s in the household as either head or spouse or

non-head or spouse (5 = hos,nhos) and r being when the individual passed away. "0th is

the number of deceased prime-age adult individuals. Equation (2) is the labor constraint

showing that time in the household is shared among leisure time( L1 ) farm labor( L fa )

wage employment (LW )or non—farm work and domestic work( Lh ) including nursing the

sick. L( Mo ) depicts that adult death affects time allocation among activities as well as
the total amount of time stock available in the household.
Equation (3) is the production function constraint. It shows that agricultural

output level( Qa ), does not just depend on quantity of purchased inputs like fertilizers and
improved seed ( Vv ) and quantity of effective family (L30 ) and hired( L: ) labor used, but
also upon mortality status (quality) of the household’s managerial skills measured by the
stock of knowledge and or farming experience of the decision maker and household

members Mg( Mo) and the stock of farm assets A( Mo) used in agriculture. Zq is a vector of

farm characteristics such farm size, implements, rainfall, and tenure that affect
productivity.
Equation (4) is the full income constraint that balances consumption expenditure

with total household income. The value of marketed plus home-consumed agricultural

. * - e o
goods must equal farm-restricted profit 7: — PaQa — PvVv — wLW plus net credit NCRD( Mo ) ,

defined as credit minus savings, and remittance income Re mit( Mo ) and other non-labor

based income Y” [i . The variable Pa is price of agricultural output, Pv is the price on

31

inputs and w is the wage for labor. If we let the right hand side of equation (4) be equal

to Y , we shorten it to equation (5) as below.

PaXa+Pme =WEE“Mo),Qa(L‘}a(Mo),A(Mo),Mg(Mo))] . (5)

Equation 5 shows the possible pathway in which HIV/AIDS related prime-age mortality
and morbidity can affect rural household. The FAO 1995, Barnett and Blaike, 1992 and
Chapoto (2006), developed a flow chart that summarizes potential pathways by which
prime-age death influences rural households (Appendix B, Figure B 1). From equation 5
we can see that prime-age mortality can potentially affect household utility from farm
production through labor loss, managerial skills and asset loss including working capital.

In agricultural household models, the household is the basic unit of management
decisions on both production and consumption (Strauss and Thomas, 1998; Benjamin,
1992). As a producer, it uses its own labor, land and capital endowment and purchased
inputs to produce agricultural commodities for consumption or sale, depending on
resource or market constraints. Under perfect market conditions, there is separability
between production and consumption decisions where the household can solve
recursively first its production problems and then allocate the full income to
consumption. When perfect markets do not exist, decisions are made simultaneously
and are therefore non-separable.

Under separability, households can find market substitutes for farm labor so they
hire labor on the market. Characteristics such as household composition and health of the
farmer, therefore do not affect output supply, farm labor demand and farm profits
because a household can hire in labor, a perfect substitute for family labor (Benjamin,

1992; Pitt and Rosenzweig, 1986). Changes in household demographics may affect

32

consumption and time use decisions but not production decisions. Production decisions
depend only a vector of prices and the production technology. The resulting theoretical
prediction from separability is that farm production decisions with respect to some
maximized profit will not be affected by adult death or the health status of household
members (Beegle, 1997). The utility maximization first order condition obtained from

the optimization assuming separability yields the following reduced form equations:

Output supply function: Q a = f ( Pa, Px,w, Mg( Mo ),L( Mo ),~Zq ) (6)

Factor demand function: X I = f ( Pa'Px'W' Mg( Mo ),u Mo ),‘Zq ) (7)

If the separability assumption holds, then the influence of adult mortality (Mo) on output

supply and factor demand is non-existent as shown below:

 

 

 

 

&0=&a*wg= ‘ (8)
awn aha; awn
(0) I”)
fa
aQazeQMa :0 (9)
obha (11h clflo
(0) (*l
i i
0% zax *oMgz (10)
¢JW0 awn; awe
(0) I“)
i i jh
cfl _c% *1. -0 (1D

obh9-6Lfa .awb —
(0) (-)

Equations (8) and (9) show that as sickness and death in a household increases,

managerial capability and family labor supply decrease. However, their effects are not

33

felt on output because family managerial skill and labor, which are household factors, do
not influence output according to the separability condition in the agricultural household
economic model. Equations (10) and (11) depict the same but for inputs such as
fertilizer, seed and labor.

Assuming separability simplifies things but it is a strong assumption to make. Pitt
and Rosenzweig (1986) model the farmer’s response to health on farm labor and profit
and tested the assumption of separability of consumption and production decisions and
failed to reject it using Indonesian household data. However, because of the restrictive
assumptions of separability, their results cannot be generalized widely. In Malawi,
farmers do participate in markets. We have observed members of the household
participating in farm activities as well as hiring extra labor. We can therefore perform
empirical models that test the hypothesis that adult death has no influence by including
these vectors in an output supply function regression model that test to see if the
coefficients are significantly different from zero.

We postulate the following three hypotheses to assess the impact of PA adult
death induced losses of labor time and farm managerial knowledge and skills on
agricultural production among rural agricultural households in Malawi.

Hypothesis 1 .' Adult death, gender and position of the deceased PA individual have no
effect on agricultural production of afllicted households.

Hypothesis 2. Human capital and labor losses induced by PA mortality have no effect on
agricultural production among afllicted households.

Hypothesis 3. PA mortality has no short-term, medium-term or long-term effect on

agricultural production of afilicted households.

34

Hypothesis 1 challenges perception that the death, as well as gender and position in
household of a deceased prime-age individual affect output supply since decisions on
which area planted to maize (a cash and food crop) versus other crops and total area
planted as a whole rests primarily on male heads of households. Hypothesis 2 tests for
the roles of human capital and labor time loss in determining production. The third
hypothesis looks at the Short-term versus medium and long-term impacts of PA mortality
on afflicted households. Conventional wisdom dictates that the longer the period after
death, the higher the likelihood of finding no effects of prime age death on afflicted
households because the households recover from the death shock with time. Hypothesis 3
allows this to be tested with a null hypothesis that afflicted households do not suffer any
effect on agricultural production in the short to long-term period. In other words, they
recover immediately following the death shock. This has implications for mitigation
strategies, for which decisions are needed concerning the appropriate period to introduce
mitigation strategies, as well as the duration of interventions, such as when is the best

period to target mitigation strategies and for how long.

Death Dynamics

The effects of adult death, particularly with AIDS are inherently dynamic. The
long-term illness triggers household response adjustments in preparation of the event of
death while the death event itself brings a shock to the households, in which resources are
reallocated towards funeral expenses. After the funeral, more death adjustment takes
place as the household goes into recovery. The decay of ex-post effects of death depends

on coping strategies available to the household. A dynamic model with time series data

35

ranging from before the death shock to well after the funeral would best capture death
dynamics from illness to post-funeral adjustments. Evans and Miguel (2004) included
indicator variables up to 2 years before death and up to 7 years after death and found
significant ex-ante death effects on school enrollment and participation in Kenya. Time
series data that enable assessment of dynamic ex-ante and ex-post effects for Malawi was
not available. The model proposed above is static, capturing two points in time, a period
prior to death and one after death. It therefore essentially captures permanent adjustment
effects rather than dynamic effects of death. As death impact literature grows, we hope
more multi-period studies on death impacts will emerge to capture more of the dynamics
of death.

The theoretical model specification of the reduced output supply function
developed above therefore depicts two static models one at t =t-1 (the baseline) and
another at t = t (the follow up) where the death event occurred in between the two points
at t = t' in some (and not all) households — the afflicted households where losses in labor
time and human capital encapsulated in PA adult death potentially affecting production
through a decline in family labor time, and loss of farm managerial skills. We can restate

equation (6) as shown below to capture the two time periods:

Q

_ . q
ait ’fIPat’th'wt’MgiIIM0"t*)’Lit(M0ijt*)'Zit) (12)

U

where subscripti represents household, ijt * represents household i with a Prime age

adult mortality of deceased individual j who died at time F . Such a two static period

36

model would most likely be underestimating the death impacts compared to a multi-

period dynamic model.

3.4 Data and Model

HIV /AIDS is a long-term disease, and adult mortality shock may last long in poor
rural economies. To best assess the effect of adult mortality on agricultural production a
long time period is necessary. A panel data set is best suited to capture these effects
because it allows the analyst to follow the same individual household across time periods
taking measurements before and after the event or policy interventions of interest. Panel
data allows the comparison of affected group (treatment) and the unaffected group

(control) as a natural experiment in social sciences (W ooldridge, 1999).

3.4.1 Data

The data set used in this study is a 13 year-panel data constructed through re-
surveying a baseline Maize Variety and Technology Adoption survey (MVTS) of 420
households conducted by CIMMYT and the Ministry of Agriculture in Malawi
(CIMMYT/MOA) in 1990. The re-survey or follow-up survey called Malawi
Agricultural Productivity and Adult Mortality Study (MAPAMS) was conducted in 2002
as part of the doctoral research fieldwork. The objectives of MVTS were to profile farmer
adoption behavior of maize varieties and to assess how household factors influenced
farmers’ selection of maize varieties (Smale, et. al 1991). It collected data on

demographics, agronomic practices, maize varietal information, production costs, yields,

37

wages, prices, income and rainfall records. Out of the original 420 households, the
Malawi Agricultural Production and Adult Mortality Study (MAPAMS) successfully re-
interviewed 351 of the original households collecting data on household membership,
mortality, AIDS awareness, asset depletion and agronomic practices data. Death history

was captured retrospectively for the period between 1990 and 2002.

3.4.2 Sampling Method and Procedure

The baseline MVTS was designed as a module attached to the Annual Survey of
Agriculture (ASA), for a subset of households included in the 1989-90 national sampling
frame of Malawi’s National Statistical Office focusing on major maize producing regions
of Malawi namely; Blanytre, Liwonde, Kasungu, Lilongwe and Mzuzu Agricultural
Development Districts (ADDS). Three ADDS, Blantyre, Kasungu and Mzuzu (see Figure
2.1) were purposively chosen to represent contrasting agroecological and household
economic zones reflecting factors hypothesized to affect maize varietal adoption and
production. Evaluation officers listed representative Enumeration Areas (EAS) from each
of the ADDS and used systematic random sampling to select 7 EAs out of 21 in Blantyre
ADD, 7 of 10 in Mzuzu and 7 of 25 in Kasungu. The survey area therefore represented
the higher potential, major maize producing areas within three ADDS. In each BA, 20

households were randomly selected.

38

3.4.3 Sample Size, Attrition and Comparison of Means

Longitudinal data are usually plagued by the problem of attrition, which normally
increases over the course of time and cause selection bias to occur. With time, households
dissolve due to divorce or death in the family. Also some households may be located but
either refuse to participate or are unable to participate in the re-interviews. Not all
attrited households have to have a death for a bias to occur, if that attrition is
systematically related to adult death, then there is a potential for selection bias. For this
study, the sample size dropped from 420 to 351. We explore the 69 attrited households

next
Attrition Issue

Reasons for attrition were enumerated in Chapter 2, on page 15. Out of the 69
households that attrited only 2 households had PA mortality. The remainder was for
reasons other than prime-age death. As shown in Table 3.1 Blantyre ADD had the lowest
number of re-interviewed households (107) and the highest attrition among the three
ADDS. Kasungu and Mzuzu ADDs each had 122 households successfully re-interviewed.
The high attrition record in Blantyre ADD can be attributed primarily to three main
reasons. First, located in the Southern Region, Blantyre ADD has a high population
density and there more land pressure. Second, Blantyre ADD also has the main
commercial capital of Malawi, Blantyre, which offers promise of urban employment and
is more likely to entice younger families to migrate in search of job opportunities. Third,

Enumeration Areas bordering Mozambique reported 21 households or nearly a third of

39

NOON m2<m<2 do coo. m>H2 ”8.50m

 

 

 

 

mm mm we K _mm 0.3
mm m: n _ _ 0. NS 03 QQ< 3:32 52:82
em cm x w_ mm mm. 03 QD< $5.8M 3:50
mm 3V mm 3 om R: 03 QD< 2552mm Eofisom
2e
3 NOON 5 A3 59$
9 5 A8 A3 Swot owe 832385 03— coir/5
5:352 :38. 2.25m 232 .255 0:0 -2 E 833585 EoEmoBEQ
ed Co ammo. 8 53, mEosowsoI wEocomso: _BB_=otw<
own .8602 mnemon— <m mo 8:on mac—9632.— ...o LonEsz Co 59:52 o3§=82mom 3:305

 

.83 use 33 5252. 32:2 3.5.. 5 35.5.5 .3 £53.85 owed—5.... mo coco—«>95 £22.32:— ue 53:52 fin mama;

4o

attrited households relocating to Mozambique following the end of the Mozambican Civil
war. The overall attrition rate of 16.4 percent reported in Chapter 2 for this study, given a
13-year interval, is by comparison in the literature reasonably low. Surveys in developing
countries of a similar nature reported attrition rates between rounds of visits ranging from
6 — 50%. Alderman, et al., 2001 (page 81) reports the following attrition rates for listed
intervals and countries: Bolivia, 2-year interval, 35%; Kenya, 2-year interval, 41%;
Indonesia,4-year interval, 6%; Nigeria, 5-year interval, 50%; South Africa, 5-year
interval, 16%; India, ll-year interval, 33%; and Malaysia, 12-year interval, 25%.

Our data indicate that 71 of the 351 retained or non-attrited households reported a
prime-age adult death, which implies 20 percent suffered PA mortality. Among the 69
households that attrited only two households had a verified PA death implying 3 percent
of those households suffered PA mortality. We can thus conclude that PA adult death is
not a major contributing factor to attrition to the extent it could introduce potential bias to
the analysis. If attrited households suffered a higher incidence of PA mortality between
1990 and 2002, there would be a major cause of concern about possible systematic

attrition leading to selection bias.

Comparing Means of Attrited vs non attrited households in initial period

A high attrition rate is not the only major cause of concern, but its occurrence in
conjunction with systematic differences between attrited and retained households. To

examine this potential cause of bias, we examine average levels of control variables

measured at the initial period of 1990 using a test of means. Results of the test of means

41

.8:onch do 63. 586m or. 623535 Co _o>o_ 2823 mi. .oocao:_:w_m Co 65. :5on _ is.

”382
Sam m2<m<2 um oofl m>92 ”Sea no 85cm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8. _ - owe. a: New 38 and £283 3.85 53:: co 33>
Ed- :3. So 2d NE 35 £288 3335.: co 33>
E; :5 .33 m8 .2 3.0 9283 23 323: so 33>
w? imam 2 .o 8.0 moo 3o 35 33 32323 335-8: Bee
23 some $5 3.0 sec N _ ._ E: 3; 323:? 332
Re. E03 Be So 5o 2; as 3; 333:3 32 EB
Em 5w _ .o N; 8.0 95 is 88-32 53.3 5.3 3... 3:5
2.0 So :5 2 .o .8 :d c3535 +8 32. 235335
3.0- 86. one 2d mac :3 23:55 +8 9.3 22. €35
: ._ E .o Ed N: 3.0 NS c3535 3 e 2 3:5 E
Ba :meo sod cod 33 w _ ._ 2335 on e 2 23. E
8.. E .o 33 Se 93 Ed c335 3 e o 2.5
42 :ch :3 8.0 33 Rd c3535 3 e 0 £5
33 wood 3o one Ed one c.5235 23E 23B
33 83 who :8 Be 35 2335 .53: 3.2
m: 53.0 :.~ N3 02 w; 23:55 3; 2338:
2: items Se 9% as w; as; .5333 2.33 3< 255
was :86 8.: 6.3. E: as? 233 as: 23325 co 36

S 3 :3 3 3V 3

”gm.“ awe—2 >0Q .Bm ado—2 >oD .Qm Gav—2

.mmnz

Buz 52.03%

oocouomfifl wEozvmzom 002.534. tear—34.x. :OZ

 

 

 

 

 

3.5a 82.5“ 2 Base: 853.335 32.32.. 83: 3...: an 2.3.

42

in Table 3.2 show that the attrited households were smaller and had younger household
heads, less total cultivated area, and fewer PA male adults and male boys. Younger heads
of households, fewer males and less total area cultivated is indicative of households that
are more likely to take up opportunities to move elsewhere. 'The results of these
univariate comparisons seem to suggest evidence of systematic attrition. Alderman et a1
(2000), from the experience of studying three longitudinal surveys showed that sample
dropout or attrition is often not a source of bias in multivariate analysis. This observation
along with a low PA mortality rate of 2 percent among attrited households leads us to
conclude that attrition bias due to prime adult mortality is very unlikely in this Malawi

data set.

3.5 Estimation Strategies and Empirical Model

3.5.1 Estimation Strategies

Estimation of prime age adult mortality impact on rural households using a
longitudinal data set is best done using a counterfactual approach to capture temporal
effects of shocks across households and over time using a difference-in-differences
estimator. In a two period data set, the death shock occurs between two points in time
creating a pre-incident (ex-ante) scenario and a post-incident (ex-post) scenario. At the
same time, not all households are afflicted with prime-age mortality. This gives an

additional dimension of “with” death shock and “without” death shock. Analyzing the

43

“before” and “after” scenario alone or the “with” and “without” scenario alone could
underestimate or overestimate the impacts.

Table 3.3 depicts results of a comparison of means on pre-incidence household
characteristics stratified by PA Mortality status. The results-show that as of 1990, there
was no significant difference in any measure of total area under cultivation between
households that were later afflicted by death and those that later were not. Households
that were later afflicted by PA mortality had significantly smaller family sizes, fewer PA
males and boys, younger heads of household and less educated prime age adults in the
earlier period.

Table 3.4 reports results of the statistical differences between afflicted households
and non-afflicted households in 2002 after the death shocks are recorded. The post-
incidence scenario now shows significantly less total land area cultivated by afflicted
households compared to non—afflicted households. This is largely due to the difference in
non-maize cultivated areas. There is also a statistically significant reduCtion in the
number of elderly males 60 years old and above but no significant differences in total

household size.

.oocmoEcwfi do _o>o_ E85: or. .oocmoc_=w_m .«o 352 E88: mi. .oocsocEwa do _o>o_ E083 _ I...

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5202
Son m2<m<2 a. dog m>HE ”330 :0 ooesom
8.7 end. dog. Nm.m and Sam QEdddL 2:02: Seduce do 33>
. vod- :dd- mod mmd mm: vmd QEdddL 30:33:33.2 do 2:5»
3..— Sud cod mmd :md mod 3:280; :3 V39mg: do o:_m>
m_d 5d :md 2d 3.0 _ do 23:55 +00 own 23:3: 332m
ovd- Nod- cmd 9d mmd Cd C2755 +00 own 23E 3:65
C: 9d wed .9: mod NN: GEE—.5 mm 8 m. 235“: <m
mdN Emmd odd cod vwd w: 23:55 am 2 m _ 29: <3:
mm: id mwd mod omd Ed C2255 E 8 0 $5
VON :de wad cvd wad Ed 23:55 3 9 c whom
vod wood Ed dnd Ed dmd C3255 3:35 23:5”:
vwd wdd wmd .vd odd wvd €38.35 ESE 232
mud Ecwd _ _.N Nmé omd £6 23:55 35 30533:
dwd ...dw.~ 36 03.0 mm.» M: d memo» cosmos—do 8:63 ow< 25.5
wvd lmdfi 09.2 Sdm £23 3.? @5quon 20:83: do owned

wezatwbokgb 3.223%
d_d wood dmd 8d de 2d 35 323 333:3 38:78: 30H
wvd- vod- Ed _ 2 mod 3.. $5 «23 333:3 0232
mmd- vod- cod on: mwd mm: $5 38 333:3 9:: .88.
85.233530 :23

8 A3 A3 A“: Q: 3

“Sm-“ =32 >oQ .9m :32 >09 .9m 532

KHZ wvmnz

ago—0:332: 33:38:

853:5 bzmtoz <m 3:552 <m :oZ

 

 

 

 

mafia mam—3.5:. <m .3 35:33. 33.3.: 5 85:53.23? 32—3—5: 83: 85295.9:— md 2:5.

45

 

033533 :0 ~32 300:3 or: 633333 :0 _0>0_ 300:3 was." 60:35:30. :0 33. 300:3 _ is."

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

30qu
50w m2<m< 2 a. dog w>F2 339 30 082%
odd Sud EN whd dm.m m: ._ 33295.5 0.3 x3303. :0 33>
om: wow vfim: New 23m 2.2 93225005 33> £033 V.3303:
mod cm: ON; mmd mm: _ 33 3320005 33> £083 team—3:.H
mm: 2.: mm: whd 2.0 No; 032000; 33> 8003 83m
Rd 2.0 Nvd mNd mvd vmd £0:8:5 +d© 0mm 0380.: 3.0305
VNN Eh—d de :d mvd vmd £3825 +00 03 038 3:035
Nwd- «:3- mo; 3.: we: om: £3835 om 2 m: 0388 1?:
5d Ed 0.3 3.: :3 hm: £3825 on 9 m. 038 <m
Sud- mdd- mo; mwd wad wbd £3835 E o. 0 2:0
Nvd mod :wd mud Nod mud £3835 3 2 o whom
and mod and hmd :d Sud £3835 8:335 0380m
dnd mod cod and vod wmd £3835 3&8 032
mnd dmd mud Nvd _w.m med £3835 0N3 20:033.:
Nod 5d OWN 36. mw.m 3d 30: 0% 030mm :0 930» :2333 :302
8.: cod 3.: is 3.: mg as» 8:825 333 ow< 255
mm: OWN 05m. 2 dm N03 mhdm £303 3303 20:032.: :0 0w<
3:33:33SU 3.55::
mod. .LNd de mmd mod cvd 35 303 333:3 0N38.:0: 35,—.
dmd «dd Ed 5d :.d vwd 35 303 333:3 0N32
ow: .mmd Ed ed: 2: _.m._ 35 30.3 333:3 30:2 35:
3.3320339 3:3

5 35 A3 A3 35 35

33-: :32 >00 .9m :32 >09 .3m :32

_>HZ dwNnZ

0203303 33:032.:

00:03:39 333302 <m 333202 (E :02

 

 

 

 

 

was“ 3.5.8:. 3. S 838: $232 5 83.53229 238.5: 388 8535.3: 3. 03.3

46

The pre-incidence and post-incidence estimates, though giving us important
information, makes it difficult to reconcile “before” and “after” with the “with” and
“without” scenario. The best estimation strategy is to use a difference-in-difference

approach as illustrated in Table 3.5 below.

 

 

 

 

 

 

 

 

Table 3.5 Illustration of the difference-indifference estimation approach
Household category Year of Year of Difference across time
Observation Observation
2002 1990
Treatment Afflicted Households
(PA) YPA02 YPA90 YPAOZ "YPA90
Control Non Afflicted
HOUSChOIdS (NPA) YNPAOZ YNPAOO YNPAoz ‘YNPA9O
Difference across time and across household (YPAOZ —YpA90) — (YNPAOZ —YNPA90)

 

categories (Diflerence-In-Ddference)

 

The variable of interest, dependent variable Yit in the reduced form output supply

function is measured through two subgroups: Prime—age mortality (PAM) afflicted
households (i=PA) as the treatment group and no prime-age death afflicted households
(i=NPA), the control group. These measurements are taken at two points in time, 2002

(t=02) and 1990 (t=90). The difference-in-difference (DID) estimator E(,6) is obtained as

shown below.

5(5): 15‘ YPAOZ )‘E( YPA90”'/E( YNPAOZ )‘E( YNPA90 )1

=E(AYPA)—E(AYNPA) (13)

47

 

The DID approach therefore allows comparison between treatment and control groups

while differencing between two time periods t=2002 and t=1990.

3.5.2 Econometric Model

The base econometric model for this study follows a similar approach used by

Chapoto and Jayne (2005) and Chapoto (2006) and is given in equation 14 below.

)Etzflt+t*Moij6+ai+,uit, t=0,l (14)

Y“ denotes the outcome measure such as area under cultivation in household 1' at timet.
The variable fit is a dummy variable that equals zero when t=0 and one when t=l. This

dummy variable facilitates change of intercept over time to capture secular trends in the

study area. Moi]. is a vector that carries mortality status variables (binary variables for

prime-adult death for individual member j by gender, position of household and also a

continuous variable of count of PA adults that died.). The mortality variable has no time
subscript because death event occurs between two-observation points, i.e. no interviews
were conducted during funerals. The death history was collected retrospectively at t=1.

Moi]. = 1 if the household experienced a prime-age adult death between 1990 and 2002,
otherwise Moi]. = O. 6 is the treatment effect of prime age adult death. The ai variable is

the unobserved household heterogeneity capturing all time invariant factors specific to

48

households that affects Y” . The term ,uit is a stochastic disturbance that changes over time
and affect Y”.

To obtain an estimate of the impact of prime-age mortality, we need to obtain
change in outcome variables over time between afflicted households and non-afflicted

ones by differencing equation 14 at t=1 and t=0 as below.

(YH —Yl'0)=(fll —fl0)+(l* Maya-0* M0175)+(ai -ai )+(#il -Iul'0)9

AYi =¢+Moij6+ Apl. (15)

Equation 15 is the result difference-in-difference estimator that can be estimated by
ordinary least squares estimation assuming that neither initial conditions of the household
nor attributes of the deceased affect the treatment effect 6 .

An increasing number of studies, Ainsworth and Dayton (2000), Beegle (2003);
Yamano and Jayne (2004), Yamano and Jayne (2005), Chapoto and Jayne (2005), and
Chapoto, (2006) have found effects of prime age adult death to differ by initial household
conditions such as stock of education, income, asset levels and dependency ratios. There
is therefore a need to control for heterogeneity by introducing a vector of exogenous

household pre-incidence level covariates Rt.0 into equation 14. The variable RI.0 is a

vector of individual households’ explanatory variables for household i at timer = 0.

This vector is largely set for pre-death conditional right hand side variables that capture

heterogeneity among households such as household 2" and farm characteristics 24 and

49

It]! indicates the effect of time interacted by initial covariates RiO' The resulting

equation (16) is stated below.

Yitzflt+t*Moij6+(Ri0)*ty/+ai+pit, t=0,l (16)

Following the same approach we obtain difference-in-difference by differencing equation

16 to obtain equation 17 below.

(Yil 'Yio)=('61 -,30)+(1* MOU5-0" M01j5)+[(Ri0)*lW-(Ri0)*0V/I+(ai -ai HUI“ fill-0).

AYi =¢+M0ij6+(Ri0)*ly/+A,ui (17)

In order to fully analyze differential impacts of PA mortality three more
adjustments are made to equation 16. First, we permit interactions of the initial covariates
with the mortality variable to examine if there are any treatment interaction effects.
Second, we permit inclusion of any other differenced control variables that are

uncorrelated with prime-age adult death and Afli . Third, we permit for control of location

or area effects. The resulting economic model is given as equation 18 below.

AY'. = ¢+ Moij6+ Riot/1+ RiO * Moij6+ ARi<I>+ LOC§+ Aui , i: l,...,N (18)

In equation 18, AYI. is the differenced outcome variable for each observation, (I) a

constant, Moi}. PA mortality indicator variable and 6 PA mortality effect. The

50

coefficient w represents estimates on initial control variables R1.O while (I) is coefficient

on differenced control variables uncorrelated with death. The location dummy variable

LOC controls for the fixed effects of area specific features ; such as soil quality and

local cultures and Al’i is the difference between errors at time 1990 and time 2002.

Equation 18 is estimated using Ordinary Least Squares (OLS), fixed effects and cluster
regression analysis to obtain the estimate of the effect of prime age adult mortality on

agricultural production in rural Malawi households.

3.5.3 Variable Construction

Outcome variables ( AYI. )

The output variable considered in this essay as a measure of agricultural
production is change in total crop area under production in hectares. This variable is
preferred to crop yield which is influenced by factors outside the farmer’s control. Total
area under production is a choice variable directly under household head’s control and
would therefore reflect his or her assessment of what labor, money and human capital the
household has in stock after a PA death. The total area cultivated is also disaggregated
into change in area under maize and change in area under non-maize crops both in
hectares. The disaggregating is important to observe effect of PA mortality on change in
maize area since this study is located in the main maize areas. Maize is an important
staple food crop in the area that it is grown for home consumption with the surplus

marketed. It therefore fulfills direct home food requirements as well as access to food

51

when surplus is sold for cash on the market and farmer can buy other food requirements
for the home. The non-maize crop area represents crops such as cotton, tobacco and
groundnuts. These are high value but high labor requiring crop enterprises. Expansion of
these areas implies crop diversification in the area. We can there measure the impact of

PA death on crop diversification in the main maize area.

Mortality Variables ( Moij )

The MOij vector of mortality related variable contains binary variables as well as
continuous variables. Binary variables (03") indicate if a household i has been afflicted

by PA mortality (1,0), of individual j by gender (j=m,f ) whose position in the household

5 is either head or spouse or non head or spouse at time t' , while continuous
variables( nDtj’S) is the total number of PA adult deaths in afflicted households such that

coefficients to those variables measure the effect of the death of one PA adult j on
average. We consider the occurrence of a PA death and then we disaggregate to see effect
of gender of deceased, position in household as a spouse or head of household or non-
head or spouse — other household member. To assess the short, medium and long-term
effect, we disaggregate the period after a PA death into four categories of time elapsed
since death. Short—term is the period 0-3 years after a death shock. These would be death
that occurred between 2000 and 2002. The medium term is the period of between 4 to 6
years elapsed since a death shock occurred. Death for the medium term occurred between
1997 and 1999. The long term is the remaining period with death that occurred between

1990 and 1996. We disaggregate into long-term 1 indicating death between 7 and 9 years

52

ago and long-term 2 indicating death between 10 and 13 years ago. Figure 3.1 presents

the death variables in a diagrammatic form.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PA death Variables
PA Mortality
[B, C, T]
l 1
PA Male Mortality PA Female Mortality
[3. C. T] [8. C. T]
Head or Non-Head or Head or Non-Head or
Spouse Spouse Spouse Spouse
[8, C, T] [8, C, T] [B, C, T] [B, C, T]

 

 

 

 

 

 

 

 

 

 

 

Figure 3.1: PA mortality variables indicating disaggregating by gender,
position in household.

Key: B = Binary variable is D,”

c = Continuous variable "of"
T = represents the term [short (0-3 yrs after PA death), medium (4-6yrs) and
long (7—13years)]

Other PA mortality related continuous variables are monetary loss, human capital
loss and average number of years since a PA death shock. The monetary loss is a sum of
medical and funeral expenses, foregone contributions in remittances and other non-farm

income brought in by the deceased during last year he or she was active, measured in

53

thousands of Malawi Kwacha. Implicit in this variable is the assumption that the
deceased individual would have continued to work and make the same financial
contribution had he not fallen sick or deceased. This does not capture the potential that
he could have been promoted. We also do not have data on remittances foregone by non-
resident family members who moved back. The monetary loss variable could potentially
represent an underestimation of the financial loss. The human capital loss is empirically
measured in two ways; the formal education years lost from the death of a PA adult and
the years of farming experience lost from deceased. The prime age adult stocks of formal
years of education and years of farming experience in a household changes due to entry
and exit from household in search of jobs elsewhere and household members growing
into and out of the prime age group (15—59). We had to control for those non-death
related prime age human capital loss or gain. The average number of ears lost since a PA
death shock controls for the time the household has had to recover from shock.
Another important mortality group is the elderly death 60-79 years old. Alumira
(2002) reports that with HIV/AIDS killing prime-age workers, elderly people are
expected to come out of retirement to help care for dependents. In cases where they may
not physically contribute much manual power, their knowledge and experience may be
critical. Indeed during informal survey discussions and formal survey field observations,
we encountered farmers in that age group still active in decision making and farming.
We included variables measuring either the total number of elderly deaths or number of

male and female elderly deaths to examine the effect of their death on cultivated area.

54

Price variables

Prices of inputs and output theoretically influence production levels in an output
supply function. We use distance to market as a collective measure of the effective price
farmers face (Strauss and Thomas, 1995). Access to market or availability of markets
has been shown to influence relative prices in empirical analysis (Acton, 1975). The
pre—death data set did not measure distance to market, so we use the values observed in
2002. Since analysis does not include households that moved away, the distance to
market is essentially the same as those of 1990. The households are assumed to be net-
sellers, so that the longer the distance is to market, the more expensive it is to produce

because of high input cost.

Initial Covariates ( Rio)

As mentioned earlier, literature has shown that the effects of prime age adult death tend

to differ by initial household conditions (Yamano and Jayne, 2005; Chapoto, 2006). We

therefore control pre-incidence level household ( Zh ) and some farm characteristics ( Zq )

as discussed below.

Household Characteristics ( Zh )

The household characteristics ( Zh ) vector of variables includes household size

disaggregated into infants, number of boys, girls, prime age men, prime age women,

55

elderly men and elderly women. Variables like infants were be dropped to maintain

parsimonious regression if insignificant.
Farm Characteristics ( Zq )

For this farm characteristics vector of variables ( Zq ) we control for total area
planted in period one to capture pre-death land use pattern. Other farm characteristics
such as soil types are captured in the location dummy variables. The rainfall variable is
treated differently. The early rain falling in October and November has an influence on
the total area planted by farmers each year. If there is a good early rainfall, farmers are
able to prepare more land compared to when there is poor rainfall and the ground is hard
to till. In our difference-in-difference model, the amount of early rains that a household
receives in seasons 13 years apart are not dependent on whether a household is afflicted
by PA mortality such that the change in rainfall would be correlated with the occurrence
of a PA death. Therefore rather than using the pre-incidence rainfall we used the

difference amount of early rainfall in millimeters.
Managerial Skills (Mg)

This vector of managerial skills (Mg) variables capturing the stock of human
capital comprising extension visits, training in specialization, and education of PA adults
in the pre-death period. Insignificant variables were dropped to maintain parsimonious

regression.

56

Location variables (LOC)

There are three types of location variables: 105 villages, 21 enumeration areas
(EAS) and 3 regions. We used 20 EA location dummies with one used as the base case to
capture location effects such as soils and cultural practices.

The empirical analysis of this essay proceeds in two stages. In the first, we
characterize descriptive statistics on the households that are afflicted and those that are
not afflicted. We test for differences in means using t-test to examine if there is a
significant difference between the change in total area planted between afflicted
households and non-afflicted households that is attributable to PA mortality. In the
second stage, we estimate difference-indifference fixed effects regression models to test

for the effects of PA mortality on labor, working capital and knowledge and skills.

3.5.4 Econometric Concerns

There are a number of econometric concerns that exist in the literature on impact
of adult mortality. We will discuss the approach taken by others and highlight that taken
in this study.

Measuring HIV/AIDS mortality

Measuring HIV/AIDS death has remained a challenge in micro level studies for a

number of reasons. First, HIV/AIDS per se does not cause death, but it weakens the

57

body’s immune system making afflicted individuals susceptible to opportunistic
infections such as pneumonia and tuberculosis, which then cause death. We therefore
can only talk of AIDS related deaths rather than AIDS deaths. Second, ascertaining
AIDS related death requires a clinical HIV blood test. This is very expensive and only a
few studies have used blood tests in conjunction with other data (Urassa et. al., 2001).
Third, verbal autopsies of medical field workers, close relatives and caregivers of the
deceased individuals have been used to determine signs, symptoms of terminal illnesses
to reach a diagnosis (Kahn et al 1999). The probability of incorrect diagnosis is high
when few information sources are used (Yamano and Jayne, 2004). Fourth, stigma is
still an issue that limits full disclosure by interviewees belonging to the afflicted
households. Those disclosing might belong to a unique group introducing self-selection
bias in the sample. Due to these prevailing measurement issues, majority of micro-level
studies used to date have used reported illness related prime-age (15-59years old)
mortality that excludes traffic accident, suicide and other non-illness related death as a

proxy measure of HIV/AIDS illness related death. This study uses the same approach.
Endogeneity of death variables

There is growing evidence that adult mortality attributable to AIDS is not a
random event but rather a result of the deceased’s behavioral choice related to sexual
behavior influenced by social standing, education and wealth. Ainsworth and Semali
(1998) and Yamano and Jayne (2005), found individuals and households incurring PA

mortality likely displaying unique characteristics such as higher education and higher

58

income. Chapoto (2006) found that men who work outside the farm and women who
participate in non-farm activities away from home were more likely to die of AIDS
related illness. Wealthy men were 1.4 — 1.8 times more likely to die than poor men.

Beegle (2003) stated that a sick person might decide .which rural household to
move into so as to receive terminal nursing, rendering death at household level a non-
random choice. In this case, death variables are likely to be correlated with the error
term in the household outcome of interest. This correlation of PA mortality to
behavioral patterns most likely influenced by economic and social conditions may
generate biased estimates of PAM impacts on household welfare outcome due to
endogeneity.

There is therefore a need to test for the endogeneity of death variables and if
found, to correct for it. Use of instrumental variables (IV) is one method to deal with
endogeneity. The challenge is finding instruments that distinguish between afflicted
households and non-afflicted households and are also not directly correlated with the
outcome variable of interest. Two approaches come to mind. The first is to use
generated instruments (Wooldridge, 2002). This is a two-stage IV method where at the
first stage, a reduced form probit model is estimated for each death variable;

P(Mojs=l)=st(X,Z) (17)

where js indicates the gender and household position of the individual, X is a set of

exogenous variables and Z a vector of instruments. The fitted probabilities from (1?)
become instruments for death variable categories. The predicted probabilities do not

have to be correctly specified; identification is achieved by the non-linearity of

59

P( Mojs = II X ). The draw back with this instrument is that multicollinearity leads to

imprecise estimates.

The second option is to find another variable that satisfies two requirements for
the definition of an IV, which are, (a) it must be correlated with potentially endogenous
variable(s) and (b) it should be orthogonal to the error process (Wooldridge, 2002).
Chapoto (2006 pp 83-86) considered several such instruments, among them lagged
HIV/AIDS prevalence, prior prime-age death, and rainfall shocks. The lagged
HIV/AIDS prevalence failed to be a suitable instrument because it was correlated with
both adult death variables and outcome variables. Prior prime age adult death was
suitable for differenced data but failed to qualify for pooled data due to correlation with
unobserved household characteristics. Chapoto (2006) successfully used rainfall shock
as an instrument for PA death. This study adopts the use of rainfall shock as an
instrument following Chapoto’s approach. The motivation is presented in the next

section below.
Rainfall Shock Instrument

Rainfall patterns can cause outward migration from rural areas as people are
driven by hunger to go and search for food. Not everyone will out-migrate. The prime-
age individuals with earning potential usually do. Living far away from family, men are
tempted to solicit sexual favors from women who are not their spouses. Young women
may turn to commercial sex to generate a living exposing themselves to AIDS. These

people are likely to return to rural areas when rains are good, potentially carrying the

60

virus that causes AIDS death with them. Bryceson et al. (2005) found hunger to be a
significant factor in increasing susceptibility to HIV/AIDS among three smallholder
farmer village communities of Lilongwe District of Malawi that were engaging in sexual
activities to earn a living. 1

Not everyone migrates with drought or years of plenty. Different age groups
within the prime age groups may show different pattern. We used rainfall indices that
show climatic variability for Malawi constructed by Clay et. al (2003) for the period
1970 to 1998 and interacted the drought year deviation from the mean with 9 death age
groups categories by gender to generate rainfall shock instruments. The instruments
were tested for relevance using Hausman test and results of the test are discussed in

section 3.6.3.

3.6 Results

The results section is divided into two sections, the characterization and the
empirical results. The characterization section gives some descriptive statistics on the
afflicted individuals versus the non-afflicted individuals, as well as the afflicted versus
non-afflicted households. The empirical results section tests the difference between PA
mortality (PAM) afflicted households and Non PA mortality (NPAM) households and
then presents regression results that assess the impact of PA death and the losses

associated with PA death.

61

3.6.1 Characteristics of Affected PA Individuals and Households

Characterization at the Individual level

Table 3.6 shows characteristics of prime age deceased individuals, and non—
deceased prime age individuals in the survey areas of Blantyre, Kasungu, and Mzuzu
ADD in Malawi. Between the year 1990 and 2002, a total of 85 prime age adults died
due to illness. Of those 48 (56.4%) were male and 37 (43.7%) female. Year range of
death confirms that more PA males died in the earlier years 1994 —1999 and PA females
dying in the later years 1997 - 2002. While not conclusive on its own, this delayed peak
for PA women is consistent with the expectation that men contract AIDS first and later
pass it on to their spouse, since men mostly work away from home and socialize outside
the home. Mean age at death is also higher for deceased PA males (38.1 years)
compared to deceased PA women, a pattern found in neighboring Zambia (Chapoto and
Jayne, 2006). Mean schooling for both deceased and the living is higher for men than
for women. Looking at 1990 statistics, the deceased individuals had higher levels of
education than those still living. Earlier studies showed that more educated people were
mobile, wealthier and more likely to have multiple sex partners. A comparison of the
education levels in 2002 shows that the mean years of schooling of the deceased is lower
than that of the non-afflicted. This is indicative of a possible shift in who is being
afflicted and a changing pattern of the epidemic.

Nearly 58 percent of the deceased PA males were either heads of households or

their spouses, while 56 percent of the deceased females were either heads of household

62

memo» mam—00:8 092on Sea :

£3» mam—00:8 owfigm ommfi ._ 5:353 3853mm

 

 

 

23232: 8562 .52

68889 2323?: Bee?

odo— odo—
ofim w.m~ 9mm v.3 Achofisomv DQ< 25.8—m
Nmm mom v.3 9mm 28:58 an? 33:53—
odm vfim CNN cdm Acaoztozv QQ< 332

SE 20.23.5me not 38:5 «oi 2393‘ 3 =§=£L3Q

2V 3

S S NOON - OOON
o. N_ 002 - 3a—
m C 002 - 32
o o moo. - coo—
TEEE: Sewn? mmzem gem:

o.oo_ odo— o.oo_ odo—
mé Om ~.N_ m6. on - mm
0.: md. 0.3 Nam Vm - mv
QNN m6— vém od— vv - mm
05 <2 mdm QC 3“ - mN
finm Yew We odm vm - m—
ace C :SSQEEQ om< Botany 3s 33%.? 22.: 3 meme? $23235 Enoch om<

odo— odo— o.oo_ odo—
N.mv NS 0.3V 0? £382: .555
2% men can Em asoawasm
33 333.33% new: 2: S QEESSEK

38.. _ .m $48 :06

.3 .5 .N 3 v .m E83 8:028 :32

wad: «8.2V
Nam 2: 9:33 58o a a? :32
83139395 3:333:
mom 3m hm wv £16m owméEtm mo 59:32
225m 222 Beacon 232 25:22:20

 

33.33 @5532 we «can .823 E flan-2:9: amass—ta commune?:o: 6:.“ tom—38c we mozmtouoauaao 9m «35,—.

63

or spouses. The ratio of deceased female head or spouse to females who are other
members of household is comparable to that of the non-afflicted (living) cohorts.
However, a comparison of male counterparts shows more head or spouses being
afflicted by prime age adult mortality. Adults aged 15-34 accounted for nearly half of
the deceased, 44.4 percent of males and 48.8 percent of females. Geographically, while

Blantyre has about 26 percent of the non-afflicted men in the sample, it has 44 percent of
the afflicted men. Blantyre also has a higher than proportionate number of afflicted
women in the sample.

Characterization at the household level

Table 3.7 shows household level characterization of afflicted and non-afflicted
households disaggregated by prime age adult death and gender of deceased. The ex-post
scenario shows a decline in household membership from 5.7 to 5.2 among afflicted
households and a gain in membership for the non-afflicted households. Households
with female death dropped in membership by 0.6 persons on average while those with
male death dropped by 0.8 persons. Pre-death, the dependency ratio was lower among
households that would later be afflicted by PA death compared to the non-afflicted ones.
The same pattern held for the post death scenario. However, the dependency ratio
among households with female PA death held steady at 1.27, while the ratio on
households with male PA death dropped from 1.19 to 0.97, suggesting the death of a PA
male could have been associated with dependents exiting the household to be taken care

of by other relatives since the main bread winner was gone.

64

NOON m2<m<2 8% coo— mEE “oocnom

A ESE § 332 80 3352:2820 + 338 E a + 56:5 .o .382: 8 Ramsey

 

 

 

 

 

EN 2. EN K 2

odo— c.oo_ ode odo—
EN - - 3N SN - - 3: 8&5
a: . - gm EN - - 4.: 82.22
m.mN - - ~.mN N.mN - - v.3 326:2
v.mN - - :N N. _ N - - _.om .8304
8:526
83 use. 883:3
:oztasc some 5 mm mo .88-- toztmac :03 5 mm mo 88-- 20:02.03
3:30 8m
@5283
_.om nAN 0.; v. _ m 0.5 ms ON 06 «52:85 no.5
0:2
de Qw— N.wN WVN ed Q: N.w md coo; 2:85 QED
Q5 88 wags—:0
mmd - - ONd _N.o - - oNd nob 338-52
35 33 3:33.30
8.0 - - 33 N _ ._ - - _ I no; 0%:
35 no.8 c5:
91 - - w _ ._ mm... - - .m._ 8.92:0 :23.
2 ._ So 5.. x: w? a ._ 2.. NS mama 55238
Wm ad Wm N.m ow Em _.o Em Bi beam

fitmen— mfimon— mEosomnoI 2:on 2:30
wEosomsom 232 5;» 295m 53> BEE< 222 5:3 038$ 53,
365:4 :02 $29.33: 86532.. BEEN =< =oz $835820: 8.23203 BEE...“ =< 85:22:20
mEosomzom BEE< €658.53 wBBE<

 

ANOONV .momkm

Boo: 2:<-xm

ism—a: .3 8.8 3:3
5 3.33.3: 633573: ES @835.“ .8 85232229 22325: 3ng amen—im— csa 8.35 8.8km hm ~35.

65

Area cultivated does not differ among afflicted and non-afflicted households in
the pre-death scenario. In the post-death scenario, afflicted households had less mean
total area cultivated with non-maize area being nearly half the mean area of non-afflicted
households. Total area planted fell among afflicted households between 1990 and 2002,
while it rose among non-afflicted households. A comparison of maize area versus non-
maize area reveals that average maize area declined between 1990 and 2002 for both
afflicted and non—afflicted households whereas non-maize area cultivated increased.
Non-maize area planted by non-afflicted households more than doubled while non—maize
area planted by afflicted households rose by less than half. Afflicted households were
more likely to be in the bottom quartile of per capita cultivated land area in 1990 than in

2002.
Test of Mean changes of Household Characteristics

A test of mean changes of household characteristics from 1990 to 2002 between
PA death households and NPA households was conducted and results are presented in
Table 3.8 below. The major variable of interest, change in total area cultivated, shows a
significant difference between afflicted and non-afflicted households. One-sided t-test
showed the decline in area planted significantly larger for afflicted households than non-
afflicted households. Among the other variables, change in household size is
significantly different from zero at 5% level with afflicted households losing household

members while non-afflicted households show a 0.44 person gain at the margin. This

66

_o>o_ oocmoEcmfi 28.3 0. 3:365... 93 :8on m 823653 3:88; _ 8:365 .23.

Noon m2<m<2 om one m>r52 bosom

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

no.0- mod- cod vmd mod 26 8.: EsflonEoE as? PEER» E owcmso
ovd 9.0 3.2 3.0 86 mm; 9:53 £53 22285 E omcfiu
mm; Nm.m 00.3 5.0. Smwm and 92 coop E 2:9 V6883— :39 E omSEU
cod :Lmd ovd mod mod mad 85 308 oNEE co: 8 3:83 «one :39 E omen—.0
02 3.0 3.0 $5. £6 36. 35 33:. 8 8:33 83 .88 E owcfiu
$6 2.0 mm: 2.0- m _.N mod 9583 20530: 5 :EEEo “.0 .353: .88 E owSEU
and Emfio 8N 3.0- main. Ed 3558 228:0: E 816m “—0 conEzc :39 E omSEU
wfim :mwd aim Sud- Nod 36 0558 ofim Eczema—om E «@530
N»; load 3.. 3.01 m: Nod- 35 8:33 33 :39 E owcmsu
8 A8 A3 8v 5 3

$03 :32 >oQ .9m :32 >oQ .3m :82 NOON

_ buZ owm HZ can cog cook/Hon 83258820 205301 E omSEU

8:80th mEozomsom 33:88:
5:35 45 533 «d :02

 

 

 

 

3.2—345: 5.3—u <mZ 6:5 age—3.8: 53c <m 5952— 33295“. Roumufim .w.m 2.5—.

67

 

change is largely driven by the adult population of the house as shown the highly
significant 1 percent significance level of this difference. The area planted to maize
drops for both afflicted and non-afflicted but the difference is not significant. The highly

significant change is with the non—maize area.

3.6.2 Results for Endogeneity Test

Using the Hausman-Wu chi-square test for endogeneity, and age-group-specific
drought shocks, we tested the hypothesis that the regressors were exogenous. We failed
to reject the hypothesis with p-values of 0.60 and above, concluding that death variables
were not endogenous in the difference-in—difference fixed effects estimation. Our
results agree with those of Chapoto (2006) who found DID estimator to be unbiased.
We therefore proceeded to conduct the regressions on death impacts without use of
instrumental variable regression. It should be noted that Chapoto and Jayne (2005) did
however, find evidence of biased estimates due to unobserved effects correlated with the
error term in pooled cross-sectional data. Our study did not test for pooled data case

since we were not interested in pooled data but the differenced data per se.

3.6.3 Impact of PA Mortality on Total Cultivated Area in Malawi

Earlier studies (Barnett 1994, Barnett and Blaikie 1992, Barnett et al. 1995)
indicated a priori that the death of a PA adult could result in loss of labor manpower that

causes a decline in agricultural production. This section puts to the test a hypothesis that

68

PA adult death does not result in a decline in total area planted and by proxy a reduction
in agricultural production. It also seeks to determine whether PA adult death results in a
decline in production as measured by area planted and determine if the existence and
magnitude of such impact is dependent on the gender of the’deceased or their position in
the household, that is as a head or spouse of household versus non head/spouse member.

Table 3.9 shows results of the impact of PA mortality on agricultural production
in Malawi as measured by total area under production. Columns A-D present results of
PA mortality variables in binary form while column E-H present results of continuous
form (counts), that is, a count of the number of PA individuals that died. Because of the
length of the l3-year interval in this panel and prevalence of the HIV/AIDS pandemic,
there were 14 cases of multiple deaths (2) per household. It was therefore possible and
necessary to assess the impact of not only the occurrence of at least one PA death in a
household but the number of PA deaths per household.

Results from the binary PA adult mortality variable shows that total area planted
fell by 0.28 hectare on average among households impacted by PA adult mortality. Upon
disaggregating by gender, gender of the deceased PA adults was not significant. The
gender by household position variables were both significant, however. Households that
had deaths of prime age male adult heads or spouses incurred a reduction of area planted
of nearly half a hectare (0.49ha) while those experiencing female prime age heads or
spouses had a loss of a third of a hectare (0.37ha). The death of non-household

head/spouse members did not significantly reduce total area planted. Continuous PA

69

 

 

 

 

 

 

 

 

 

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

.NNO. .0I00. 6. 00 G . .00 a . .0. Am _ 0v 3 _ .0. A: .00
m .00 m000- 0000 000.0 . .00 .000. 0000 000.0 m...w .8 oz
Sum-.0 A00. 3 NM... .903 Sew-.0 2...: Sum... 30...
i...0 *.0.0 30.0 ....0 .020 .N0.0 *0..0 .v..0 £27807.
3...: 2.»... Ann-.0 $0.0 ANN-.3 Ch... 8%: 9m...
.82.- Sod- wood- wood- .80...- Sod- wood- we..- 22. 22.38.. .0 a...
$6.00 2000 :00. 20.00 .wm00 $000 @000 80.00 I
n00.0- m .00 . .00 . .00 0000- o .00 000.0 N .00 £30 9. Sci Emu» uwa.o><
Ame-0. 2N... AN... 30... No.00 SN... 8.... $0...
. . .0- 00N0- v0.0- 0>.0- 00. .0- 030. 00. .0- 0w . .0- m... 300 Nteam.
Sure: 333m
:00. $00.
as!
:00 0.0.0 3:00.08; :2: v.25”.
$00. 2.0.
o. .0 No. .0 3.8.0:. 632.203.. 2:. 22>.
.00-.0 $0...
$3.0. $3.0. 9.3.0:. 8:09.08; 0.25".
.008 $.va
2mmv0- 3.930- 0:55... 8:00.083; 22>.
SN... .5...
2N0- 0.N0- ...£.oE 9.25.. <0
6m. .0 Q...
NON.0- m0. .0- 59.2: 22: 5.
0w. 0 .3. 0
«v0N0- .05N0- 5:85.: <n.
3:333 .30.. mmw 6E3.
.-. O n. m. D U m <
.thK hBQS=EZQU .3. ”Sad-IN?» meUtQEVw

 

35 E 0290.30 3.0 .80. E omSfiU

 

ES... Eusm 2.. 83:...5. tastes?
02.. E 0083.30 .3... .32 E uwcmzu

 

 

8.3.300

 

 

N00N. 003 £23809. ou:o.o........E-8:o-5.Eu .e 8.38. A“... .323... E 0395.3
3.3 .83 no 08.380 2.. .0 5.52. 0.5.2.5“: 0.2.82... .25 5.2—ow 5:385 «N .0 Sam—E 2E. 00 9.2.5...

70

-.:00_-..:w.m .0: mm? 3.00.. 3.00.0 =0 w:.>.0>:. 0.9.35. 00:2. 33:00 3.30:0.
E 03.00 0:000 0.03 0... 80000.. 00-00 000% 0:. :0 0.: 00380.0 0.303. 30% 0E. .00.. m. 0... 0.6 03:00 .05. 0.0:. 52. 00 500.0 0... 0.3

00.0.0008 :0..0.0..00 cm... 00 0000000 02% .3 00.000.09.30 .0: 0.03 05000 3.02m. 4.0. .0 .:00_-..:w.m .1... 60m .0 302-.330 .I. 300. .0 302-.330

0000:3030 E 00.3.3.0.-. 00 003.» 03.022 ”00.07.

98>...m NO0N min/0.2 0:0 000. m>..-.>. ”00.:0w

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

£388: _ um n 4
N000 mvd Nv-0 N30 Nvd 90.0 .30 find 00.0:mm-m
m.-0 inhw-m #0.. *wN.m mmd 30.0 v... 1.5m 3.8.52 :0 .mo-H-n.
00> 00% 00% 00> 00> 00> .0? 00% m0.EE:0 <m. :00000...
A00. .0 $0.00 30.00 20.00 A00. .0 A0w00 9.0.00 30.00
200.0 vmm-o 0wm.0 Nmmd 0.0.0 mfimd N0m-0 m0m.0 30.0000
50.3 :0.N0 A00. .0 A00. .0 :0.N0 A0l0. .0 Am 0. .0 A00. .0
:0N00- .:._0N0.0- :0N00- :0N0.0- :0N0.0- :0N00- in N00- 10N0.0- $5.0 .00—.00. 0. 00:03.0
8N0: AMm00 3.0.0 3.0.0 A0..0.0 A0000 AN. 0.0 AnNd:
5:..000- 31000.0- *immhd- 33000.0- 2:..0000- {EVER-.0- .immhd- .1.me0- 800.0 00.03.30 00.0 0:0...
A00.N0 AGO-m0 30.9 A00.N0 .00-N0 Am-0-N0 A00.N0 20.9
I..v.0.0- 22.300- £0,300- 8130.0- 10.300- 31.30.0- 31.0.0.0- ..2300- AEE0 Em. >oZ . .00 E 0w:mr_U
Sum-00 $6.00 $0.00 $0.00 @000 Amw00 Sum-.00 3000
000.0 m00.0 000.0 W000 m00.0 m00.0 m00.0 000.0 3.000 2:0 300% :0000000
€0.00 AWN00 3900 3900 $0.00 AMN00 03.00 36.00
NN..0 .0000 . . .0 0. .0 0. .0 0.0.0 N. .0 N. _.0 003:. 3.00.0 .00 07.
3m. .0 0.0. .0 .0m..0 8m. .0 .mN..0 $0.0 $6.0 QN. .0
m0N-0 CON-0 .0N0 :Nd mmmd m0N0 0.5.0 va0 00.00000 3.00.0 00 oZ
Avw00 00.00 A0000 AVG. _ 0 Suw.00 $0.00 A0000 A .0. _ 0
m00.0 .000 000.0 0w0.0 m00.0 000.0 20.0 000.0 3.0:. 012:0 OZ
A. .._0 A. .. .0 80.00 20.00 Am . . .0 A00. .0 $3.00 €0.00
00. .0- m0. .0- 0w0.0- mw00- 00. .0- .0. .0- «00.0- ww00- 00.00.00 mm 00 OZ
m < m < m < m <

 

 

 

 

 

 

 

 

EEK 0:05.009 E SEE-Ex. 0.00203...
030 E 00.03.30 00.0 3.0. E 030:0

 

EEK .0050 5 0030.23. 3.0203...
02.0 :. 00.03.30 00.0 3.0. E 00:20

 

0.0.5.000

 

 

 

00:03:00 06 0.23.

71

mortality variables, columns E to H, show a similar pattern of results, but with
households experiencing PA mortality showing an average loss of a fifth of a hectare
(0.20ha) per each prime age adult death and 0.43ha per male head/spouse and 0.37 per
female head/spouse. Mortality variables for the existence and number of PA deaths, and
the existence and number of each gender by head/spouse position were significant at 10

and 5 percent level of significance respectively, using the F—test.

3.6.4 Impact of PA Adult Mortality on Maize and Non-Maize Area Planted

Given that the study is a reconstructed panel from high maize producing areas, it
was important to disaggregate total area planted into maize and non-maize areas planted
and observe the impact on maize versus non-maize areas.

Table 3.10 presents the results of the binary variable forms for the maize area
planted in columns AD and the non-maize area planted in columns E—H. Results for PA
mortality impacts are fascinating. Existence of PA mortality in a household is not
significant factor changing maize area planted; however, the number of elderly deaths 60-
79 years old is significant. A possible explanation for this result is that the elderly poeple
who were in the PA age-group during the peak of President Kamuzu Hastings Banda’s
maize production “Chimanga Ndimoyo” policy, were still heads of households who were
influencing decisions on maize area planted due to preference for maize and the wealth of
knowledge and experience they have in maize production. A death of these elderly

would therefore result in a decline in maize area planted. This explanation is consistent

72

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2&8 A88 58 E .8 20.8 2&8 5.8 5.8
wood .85 wood wood 53. _ op. Sod Sod Em “a 35:52
so. 3 8w. : so. : Go. : g8 35.8 6&8 308
Bad $8.0 18d ... _ 8.0 $3 mod Sod :50 :8 be 5552
ab: 89: so. 3 A3.: $8 $8 $98 $6.8
$8.0- mood- $8.8 wood. 83. 53. mood- Sod- 93; 223.8; .5 o <
E8 A58 A88 58 :88 Ga: E8 E8
Bod- Sod- Sod- Sod :50 05.0 05.0 80.0 ago a 82m :8 o 202
E8 38 $8 318 aw. : 3 _ .8 so. : 8w. :
Sod Rod cad Sod 5 _ .o- 386. 3mg- 18.8 a 2-8 :55
A38 $8.8
000.0- Nw0.0 muoDEoE 550 Bag—um
A88 cum:
N 5.0. 83 2352: 5:5 2“:
so. : 858
©_N.0- mN—dn umsommhumu; ENC—om
SW8 pod
wood- £838 38323; 222
AR. 3 a: .8
Q _ .o- w _ ed- 22.3 E
808 E .:
v8.8 3. _ .8 22. E
Gm. : 608
3.2.0- :: .o. 355:. E
m o m m a u m <

 

 

 

 

 

 

 

53k 535% 5 $3313 5:32:

 

 

 

33*

 

$2 333.30 832-52 :28. E omcmco _

 

0.: 3?. 331.30 BEE ESP E o EEO

 

 

 

.83. 33 .ncmmmonwou egosguéwoosuuogu mAO «a 33mm: .35 393 6895.3 «fins—.5:
can SEE .8 @9383. 2: mo :82ch amp—E382: 2.58:2— EE .595» £33.88 «fl .3 8.3:: 23. 3d 933.

73

9&— 3 ~500_=wmm §n a anficwfi .2. 352 a Emoficwfi _.. 6805580 E mosmcflm . .8 2:? 229:?

5802

935.5 Noon m2<m<§ 0:“ 000— mEE u850m

 

 

mcozntomno Em HZ

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mod mm; 0v; “*de 5.0 .500.“ 5.0 no.0 LESS—2 :0 52¢
3.». mo? we» mo> 3% m8, mu> we? moEEEzfl <m 5:80-—
CN. 3 2m. 0 Ab _. .0 A: . 3 0.0.00 0.50 Gwdv A300
womd oomd owmd wad v3.0 05.0 3N0 mvmq E3200
20.: 8.0 $0.: 20.: 60.: 3%.: 3w: 30.:
000.0- 050. 0 0.0- 000.0- * N00- 35.0- .080. 08.0. :50 Bias. 9 3535
2 09 50.8 A 0.9 $0.0 A m; C am; 0 A51: $0. 3
$33.0- 3,102.0- *inbmfo- £030. .23.. N00- ***000.0- 1.3—0.0- Iimod- oo-ow E 08523 «En :33.
3.: A3.: 3m: SVEV $0.8 80. .0 20.9 80d
m00.0- 80.0- 80.0- 80.0- $500.0- "IL-00.0- $500.0- 1.500.0- AEEV ES >oZ - 50 E omSEU
0w. C 8v. 0 Amv. C ADV. C A300 ASN.8 A300 A300
000.0 wood 000.0 008.0 80.0- 80.0- 80.0- 80.0- 33% mm 00 Ban» .808va
A300 Cmdv 8mg 8mg Aomdv $0.00 3mg Gmdv .I
$0.0 wm0.0 08.0 3.0.0 30.0 30.0 03.0 30.0 33:. 3520 00 60:52
$5.00 Q00 80.00 2000 8:0 $0.00 Ammd 3%.:
30.0 30. 30.0 v.30 5.0 N30 :30 of .0 33:8 3520 00 82:52
@000 8410 Awmdv Ammdv Gm. C 3v. C 3v. 3 80:
20.0- a. .0- w .00- N 5.0- 05.0 2.00 $0.0 E00 3:2: <m 00 32:52
$5.00 2 .00 $000 $0.00 80.00 $500 $5.00 30.00
30.0- >3 .0- $0.0- 08.0- 800- $0.0- 90.0- 50.0- mug—SEE in 00 SnEsz
I O n— m n. U m <
Eek .035 5 336.29. b.0883 {l

 

 

m2< BEE—=0 BEE-:02 :28- E o EEO h

0.0 m2< UBm>E=U BEE EEO-- E owEEU

 

 

00:50:00 0nd Saab

74

 

with field observations where elderly people up to late 703 still attend development and
farming groups and make decisions at community levels. Indeed, some still work the
fields but in a limited sense. This has implications for extension service training and its
role as facilitator of knowledge transfer. The PA male head or spouse variable is
significant at 1% leading to a decline of slightly over a third of area under maize
production.

The non-maize area (column E—F), present a different picture from that of maize.
Households suffering PA mortality experienced a 0.16-hectare decline in area planted to
crops other than maize. The variable for number of elderly deaths is not significant.
Diversification of rural agriculture came about after 1994 with the new democratic
government of President Bakili Muluzi. The shift to non-maize production is likely to
have been embraced and driven by the PA group. A death of a PA individual is therefore
more likely to impact area planted to non-maize crop than the death of an elderly person.
Households that had a PA death suffered a decline in area planted to non-maize crops of
0.16 on average in the high maize production areas of Malawi.

Results in Table 3.11 from the continuous variables echo the pattern of the binary
mortality variables, although the impact a PA death has borderline significance (p=O. 104)
of a decline of a tenth of a hectare per deceased PA. Disaggregating by gender and
gender by position shows no significant decline in area planted.

Based on these results we reject the hypothesis that PA mortality does not impact
agricultural production. Evidence from Malawi’s high maize production areas shows that

prime age death leads to a decline a area planted with the death of male heads or spouses

75

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Avm 00 $0.00 3.00 8.00 Amlod AN~00 $0.00 €0.00
00 0 N00 0 30.0 m00.0 .00. 0 000.0- N00.0 .000 2Ew .o o:
30 .0 3.0.0 $0.: 80.0 .30 $00 $0.00 .800 In
... 50.0 *5000 150.0 300.0 3.0. .300 Wmlod 0.0.0 9.8 .o o:
85. .0 A00 0 .00.: $0. .0 €5.00 .0 v.00 $0.00 $0.00
*0000- .00-m. - ...0000- $000. 80.0- N000- N00.0- ~000- cao; 0.2.839. .0 o m
25.00 .50. 0 $0.00 £0.00 8.00 2m. .0 $0.00 90.00
500.0- .00 0- 000.0- 0000- N000 0.0.0 . .00 . .00 530 a ooEm 28 o 80>“
.5500 can 00 3‘00 €0.00 .00.: 91.8 9.0. .0 8.0.:
.00 0 500 0 90.0 000.0 m0. .0- **m0m.0- *Nvmd- $310. m; 05.00 5.80....
3.08.8... ESE-0
.300 va0
000.0- 000. EonEoE .200 EmEu.
50.8 .90
N50. em. .0 33:00:. .050 v.9:
£0. .0 80.00
0N~.0- 52.0- 3:0 m>002. EwEo.
.0500 30.3
@00- **0Vm.0- omsommBmoc EmE
80.0 .5000
V5 .0- N80- BmEo. am
20.00 .mm. .0
0w 0.0- 310- BE: mm
.3. .0 A. ...0
8:0- god. 325:. mm
0.0.3er 03.3.. e < 23$
: o n. l m o o m <
ES... 3.025309 3 $30.25. 3.033....
36.530
02.0 08¢. 08030.20 BEE-:02 .20... E omcmzo 020 «be 03520.30 3.0.2 .30... E 90:20

 

 

N08- 000: 50.39.09. 85.850.5-853t... 040 00 3.39. .300 mac...“ 0835.3 BEE-:0:
0:5 3.2: :e 0.2.9.50: 2.0 E :05ch .amsenm .5 0.508.. 0:: 305% .3 5.3.3:. <m .00 3.2—E. 2E. :6 9.05.

76

a0. 3 0:80.56 .22. 5% 3 0:85:90. .1, 300. E Ema-0.3%,.- * $8055.30 :. 85080. 0 ..o 20.? 23.800.
”3.07.

whotzw ~00N m2<m<2 0EN 000. mg: “3.500

 

00:50:00 :6 03:9

77

leading to a larger decline than that of female heads or spouses. When differentiating
impacts between maize and non-maize areas, PA mortality affected non-maize area
decline while the deaths of those in the elderly group 60-79 and male PA heads or
spouses largely drove the decline in maize area planted. '

The next section presents the results and discussion of the analysis of the impact
of PA mortality on agricultural production induced through the loss of money, knowledge

and labor.

3.6.5 The Role of Knowledge and Labor Losses in Changes in Area Planted by

Rural Malawian Agricultural Households.

Table 3.12 presents the impacts of PA mortality on area planted in Malawi as
disaggregated by losses of money, knowledge, and labor. Column’s A-C present results
of binary PA mortality variable while column D-F present the continuous form. As
detailed earlier, monetary losses are measured in thousands of Malawi kwacha while
farming experience loss is the sum total of years of farming experience of the deceased
PA individuals in each household. Results indicate that monetary loss is a significant
factor in reduction of total area planted, with a total reduction of 0.17 hectares per 1000
Malawi Kwacha lost. When disaggregated between maize and non-maize area, the
monetary loss variable was not significant for non-maize area but very significant for
maize area. The maize area planted declined by over one fifth of a hectare (0.22ha) for
every 1000 MK lost to illness, funeral expenses and forgone income whose opportunity

cost was at least partly invested in crop production.

78

.3030 00.08 :. 00000 :00... o. 96:. m.<n. m0 000000000 $08.00 .0 $0...

3.0. m... wEEE. 00.0.0. £000.52

 

N00N. 000. 50.00050. 00:0.0t.0.:..00:0..0.....0 01—0 ..0 3.000.. h050.00,. :. 00.0>...:0 00.0 00.05.:0:
0:0 3.0.: 00.02.30 00.0 .30. ..0 5:0... 0:0 $02305. ...000. .00 000. b.0338 0... ..0 .009... 0:...

8.0.000

afin 030,—.

79

.m. ..0n0 0000......w... 02.00.32. + 30. .0 ...00....:w.m 1.... 20m .0 00000.56 3. £00. .0 3000.5... ... 00005020.. =. 300.006-. 00 0:.0> 03.9.00.

3.0.02

mautzw Noon m§m<2 0:0 000. was. ”00.:om

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

—WM"Z mcomufl>h0mfio
mNo omo .00-o MNo owo .00-o 00.0:mm-~.1
NW. “3.30900 inmo.“ mo. ***mV.v 3...». .m .9535: :0 .00.-n.|
mu.” w0> mQLV m0? m0> m0? women—3D {m— £05500)—
QN. .0 20.8 Aww00 CNN. .0 80.8 Aowo0
.Nmo VONo mmmo 0. mo .cmo ammo E00300
A30. .0 AN..N0 3.08 2.0.0 A. ..N0 Amm.N0
v.o.o- immoo- **mmo.o- m.o.o- **Nmo.o- **mmo.o- $5.0 .00—.0... o. 00:90.0...
Abw.N0 SN. .0 30.00 Aww.N0 Am”. .0 Aowd0
“Isl-M.,? ***.mo.o- ***vo\t.o. ***hm..o- .303... no.0- *00vcho- 00-0w E 000030.00 00.0 .05....
A0000 ANN.N0 AVN.m0 30.8 A N.N0 AmN.m0
Nooo- **woo.o- ***m.o.o- Nooo- **w 5.0. ***m.o.o- AEE0 ...0. >07. - .00 0. 0wc0LU
Q-mo0 Amoo €0.00 A0900 30.00 30.00
0000 moo. hood wood mood wood 3.000 (A. ..o 0.00 :0..0o:0m.
Awwov A500 A.\..o0 30.9 030.00 A0000
000.0 000. N0. .0 000.0 Nhoo mm. .o 00.05 3.00.0 ..o o:
n. m. D U m <-
030 A000 002 .05 00.2.. A05 A05 00-3. A000 00.<
ONmNEi—hoz DNWNE "QuOr—i DNWNEICOZ ONMNE _~qu.—r
Eek 0:035:30 2.. 0030.23. mtg-5:... rick 0.03% ...0 0.03.0.23. b.0005}? 0000:0300

 

 

 

00.00.00 00.0 00.0.0-5: 0:0 00.0 00.0.: .020 .000. E 0 50.5

 

 

 

0.5.0 2.0 0.0.0...

 

80

The human capital variable, number of years of farming experience lost by the deceased
PA adults, was marginally significant with a p-value of p=0.l 13. Each year lost resulted
in a decline of nearly 0.01 hectare of total area planted. The labor variable was not

significant.
Discussion of the results on roles of monetary, knowledge and labor loss

Results of the role of monetary loss induced by PA morbidity and mortality are clear—cut,
however, those of labor time and knowledge and skills raise some critical measurement
issues. The major issue is the challenge of disaggregating the brawn from the brains,
labor contribution and knowledge and skills contribution all embodied in one physical
deceased being. Human capital was measured through a number of variables, including
stock of formal years of education lost, number of years the deceased was engaged in
providing specialized agricultural skills like pest control or nursery management, and
years of total farming experience. None of these were highly significant, although, the
total farming experience (years) was marginally significant. Normally, the best way to
measure human capital is to test what they know; in this study that was impossible given
the individual was deceased. We could only rely on recall data on the involvement of the
deceased; thus under or overestimation is possible given the long recall period. Second,
though marginally significant, the observation that experience was significant and not
formal education is in agreement with literature that shows in rural agriculture,

experience is more important than formal education (Yang and An, 2002).

81

The labor variable also is other measurement challenge. A better measure could
have been the hours or labor time contribution of the deceased. The recall period was a
challenge to achieve this level of accuracy in measurement. As presented in Table 3.11,
the count of the deceased PA embodies many factors that render the variable ineffective.
Despite the limitation we do still reject the hypothesis that human capital and monetary
loss do not significantly impact agricultural production in rural Malawian agricultural

households.

3.6.6 Short term, Medium-term and Long-term Impacts of PA Mortality on Area

Planted among Afflicted Malawian Agricultural Households

Table 3.13 shows results of PA mortality disaggregated by position of household and
period of death. Model A uses a mortality binary variable where the 0—3 years after a PA
death is short-term, 4—6 years is medium term and 7-13 years is long term. Results show
PA having an effect on total area planted of 0.32ha decline on afflicted households. The
effect wears off in the medium term and is negligible over the long term. Model B tests
the impact of mortality timing disaggregated by gender. Again significant impact is felt in
the 0-3 years after death and is large and significant with a male death. The death of a
male PA leads to a reduction in area planted of 0.44 ha. The death of a PA female is not
significant. Model C, in addition to considering the gender, also introduces the position of
the PA as either a head or spouse of a household. Results show larger impact from the
death of a male head or spouse and no significant effect on the death of a female head or

spouse. The death of a male head or spouse resulted in a decline in total area

82

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Em 0.00 m. I o. .00. 0... :. ...000 0.0.:0. 0 0:0 0.0E 0 00.. 0.2.3.5..
.3...
0mm..- 0.00 0 o. v .00. o... c. .0000 0.0E0. 0
80. .0
.30. 200 o o. v :0. 0... c. ...000 0.0.: 0
20.00
wood ...0... m I o ...0. 0... c. .0000 20:5. 0 0:0 0.0... 0 00.. 0.9.3.5..
.560
EN... 0.00 m o. o .00. 0... :. 5000 0.0.5. 0
8w. .0
5.00.0. 0.00.. m I o .00. 0... E ...000 0.0.: 0
830...... .0059 05. 00.0.5. ...qu .3 9.0.3:. S.
30.00
08.0 33. I 0.3.0 0.00.. m. o. 0. .m0. 0... c. ...000 0
8200
03.0. Goo. I moo: 0.00.. o o. n .00. E ...000 0m
.0... 0
3m. 0- 800. I 80.0 £00 0.32.... o... E ...000 0
.0... 0
... . and- .88 I 0008 200 m-.. .00. 0... E ...000 0
330...... 00...... ...qu .3 .5000... «Q
U m <
0.9.3.2...
:. :o...8.. .0380 .o 30:00 3.0.030
920000.. 0... .0050 .3 0o..0n.
3 BE. 530 5.3a E .55.. 580
E 5:352 <. 5:252 <. E 0:282 <.

 

 

 

.05 02:05 00.

 

< .08. :. 0wc0..0

 

 

 

.035...

00.0 .33 :0 5000 ..o 00...... 0:0 .0050 .3 0.2.82... :. 50.8.. .3 5.03.2: A... ..o .009... 2.... gum 0.00...

83

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A0... 0050.... 0050. .08. E 0 :0...0

 

 

 

 

Sod- 0 0 0.00 o 00 b 503.00 00.5000 ...000 0000 0002. 0.0.: 0
and.
womd 0.00 e - 0 .00. 0... E 5000 0.00.0. 0 0:0 0.0... 0 00.. 0.2.0002.
8.0.0.
0 .06- a» o 0 v c0050.. 00.5000 ...000 052.0%
A .. V
N u. .- 0 0 0.00 0 0:0 0 503.00 00.5000 .0000 00:0 0000.. 0.0.: 0
3.0. .0
no We- 0.00 m - o .00. 0... E ...000 0.00.0. 0:0 0.0:. 0000 0000.. 0 00.. 0.2.0002.
0N...
0.0.0- 0 0 0.00 m .00. 0... E53. 00.5000 .0000 00:0 0000.. 0.0.00. 0
00.9
.2. cm... m .00. 0... ...0... 00.5000 ...000 0000 0000.. 0.0.: 0
6nd.
Smd- g m. - o. 00.5000 .3
5.0.
00m... 0300.. m. I 0. 00.5000 ...000 0.08 0N
no.8
mm... - 0.00.. 0 0. 0 .00. 0... E ...000 20:5. 0
A0 .8
0.00.0 0.00.. a-.. .00. 0... E ...000 0.0.: 0
0030.5;
.00....02. .0002? .0 300000.. .000 .00:00 00.200. ESQ .3 b.0920... 0i
0 m. <
0.9.0002. E 50.02. .00:00
.00:000 .0 0.0.0002. .3 02.00.
.3 .00:00 .3 02.0.”. .0000 .3 0000... £000
...00n. 3 0.0.52 <n. 0.00.02 <n. 3 5:0..02 <n.

 

 

00:00 26 030...

84

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2...: A38 3.8
E .o- Sod- Sod- 8.2.5 am 8o 8
6 .8 A28 58
«.9 .0. So 890 2% Mo 2
398 God so:
:83 :53 :33 835° 9.
an: a“: so:
83. 8. .o- 83. use; 2238; 8o 9 a
3.8 9 .8 a _ .8
_m .0. 0.3 .O- cmod- ED 955 05-00 wBMEQ T535 Uomaooov 50 50:5:
3“,: a .: an:
5 no- ~88 Ed- 20 :8 goo mama 22%v 838020 Es:
Ewe 48.8 8N8
mg. 30.0 V8.0 Emu—v m 005w mane o u5><
5.0
mcmd- a» E Q 2 52533 3:580 :83.» 8:0 £93; 012:8 a
A88
mmod o a may m“ Ucm O_ =00>>5£ 63.5000 Sam—u 525 $95: 29: a
$8
>36- 0 w 980 o um. N. 205353 08.5000 Egon 330 meo: 5?th m
u m <
Begum—...0: E .5950
:OEmOQ Emsomm .5 arr—3:5: an totem UOEQL moumtm>oU
E 5980 E carom Each— 3 580 3
53a 3 £352 E 5:25: 5 3:352 §

 

 

 

 

G5 @255 32 39 E owcmnu

 

 

 

3.80 mud 28¢.

85

..E an EEE»: 1.... 3% a 285:»; i. use a 2.85:»; ,.
.mowuficofim 5 8.358.“ a we 2:? BBoB< 6802
88 m2§<2 Ea 82 m5: U850w

 

 

£288: $ng

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

3.0 9. V 96 val
*mh; mwd mm; %:02 :0 32d]
we». 3% 3% 3:555 2m: 5:801
Go 3 80.8 Ga 8
oovd mmmd ommd Esmcou
Rm. 3 8w. 3 30. V
*Rod- *Rod- *wmo. - AEVC .832: 8 853%
$0.3 $0.03 2 _.o v
532.. 55220- 335. - oodw E 383:5 «Ba :38
A wd Ammdv Gaza
...m o.o- ...Eod- 530.5- AEEV 53 >0: - 60 E o :20
33.8 2.58 30.8
wood Sod 50.0 333 a 80 £3 cosmos—go
698 80.8 35.8
~26 v2.0 $3 3:2: :02» ..o o:
36.8 A2.: A2.:
3.6 $2. .ovmd 3:25: .520 go 0:
Go. 3 $.99 $0.8
50.0 08.0 n56 a
U m <
20525: E 5250
action 830% 5 3530; .3 32.3 totem 383.530
.3 $9.00 3 wow—om Ewen .3 ESQ 3

58m E 3:38: E Ego: E 2952 §

 

 

 

 

 

95 coEaE «oz :38 E o :20

 

 

 

3:0 2...” 03:8

86

planted of 1.2 ha among afflicted households in the short term (0—3 years after a death)
and was not significant in the medium to long term.

Table 3.14 disaggregates area planted between maize and non-maize area planted.
Models A and D show that PA mortality in the 0-3 years following a death shock
significantly reduces area planted to non-maize crops by 0.23ha while no changes occur
in the maize cultivated area. In the subsequent medium term (4-6year following a PA
death) afflicted households reduced maize cultivated area by 0.33 ha. In the long-term
there was no significant reduction in maize area cultivated in response to PA death. When
gender disaggregating is introduced in models B and E, male deaths produced significant
decline in area planted to both maize and non-maize area while female deaths did not.
Death of a PA man 4—6 years earlier reduced maize cultivated area by 0.52 hectares while
a PA male death 0-3 years earlier reduced non-maize area planted by 0.28 hectares. The
death of a male head or spouse (model C) has a highly significant effect on maize area
planted in both the short-term and medium term. Such a death is associated with a 0.81
hectare reduction in maize area planted in the 0-3 year period after death and a 0.53
hectare decline in the period between 4-6 years after death, For non-maize area planted,
PA mortality is significant only in the 0-3 year period, reducing non-maize total
cultivated by 0.38 hectares.

These results show us that afflicted households are very vulnerable in the period
0-3 years after the death of a male head or spouse of household. Maize area is
particularly hard hit when a PA male head or spouse dies because there is reduction of

area cultivated in both the short-term and medium term period. It appears that as a

87

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A000 20:8
20. 3.0. 00. 0.80» o 9 v .02 0:0 E 5000 0.0:. a
00 H. 8000
00 .0- 0A _.0 w; mlo 0mm: 0:. E 5000 2:800 0:: 002: a 0:: 03:09.5:
3.0: 02.00
wlm .O- wN0.0- made m on O 3.0— 2: :0 £060? ENC—ow a
AS. C 00.00
.80.“. - 02.0- as; m - o a... 05 5 £80 22. a
000:0 ES .0250 £000. 0 0030.000; 3.05.82 <0
A “~00 00000
0 00.0- 3: .0 080.: m_ 8 0. .02 0:0 E 5000 m
30.8 A :8
000.0- 0 ”0.0- magi 0 o. h an. E 5:00 am
€0.00 30. S
80.0 ammo- as: f. as 2: E 580 a
60.8 A0000
.1. _ mmd- 30.0- 080 m E: 0:0 E 5000 m
VQNKNQ £0603 >Q MMNQUTG.) 3:00.00: 3%
0 m D 0 .0 <
2050200: U-OSOmDOr—
E :oEmo0 E 8280
03:25 03:25
8 050000: 8 020000:
0: 80:00 80:00 .3 80:00 80:00 00:08:00
0: 00:00 .3 00:00 00:00 .3 00.:00 .3 00:00 00:00
580 580 .3 580 0 530 535 £80 0
0: 0:382 0:982 0:882 .3 0:882 .3 0:882 0:882
<0 <0 <0 <0 <0 <0
25
03—02% 002 ONES-=02 E umcmcU fi<$v “50:0?” 050< DEN—2 E owcmr—U
08:20
00.3 00008.5: 05 SEE :e 5000 00 02.50 0:0 80:00 52:80 .3 5:38:— <0 00 0050:: 2:. 00m 03:0-

88

08:030.:
0:: E 80600
03.5%

8 02000000
3 80:00

3 8:0:
530 .3

80:00
E
0280
£80
.3 .0
5952

0280
:80Q
.3 a
£382

0::
E 80800
08.5% 8
05000000
.3 80:00
3 8:00
5:09 3
0:382

80:00
0
0280
£30
.3 .0
.2382

0080900
025:

£80 .3

0:382

 

0023.0 00.5

0002—75: 0:0 050.: :0 .0000 .8 0380 0:0 80:0“ .5580 .3 00:88:. <0 .8 800:: 0:0. 0:5 :0 0300.

89

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

9&0 $0.00
09.0- 0N0 0m: 0800 2 00 0_ .8038: 00.8000 5000 08E“: 0
80.00 3000
3 .0- 2: .0 0mm 0.» m: 0. 0: 000.580 5000 00:0 0000: 0.9: 0
02:00 $0.00
0 .0- Fwd- 0m: 0800 0 0. 0. 0008000 5000 00:0 2000: 08:80 0
0.0.00 03:00
0 0.0- 000 .0 mwm a» 0 2 n 0008000 5000 00:0 0000: 0.0.: 0
90.0 0300
m8: 9&0 080.0 0 Iv .00. 05 E 5000 00200 0 0:0 00:: 0 00: 0.28000:
$00 @000
m: .0- 03.0- 0 0 080 0 2 v 02880 5000 00:0 2000: 0.020.: 0
80.00 80.0
on: .0 v5.03.0. “mm m; 0 2 0 002800 5000 00:0 00:0: 002 0
30:00 fimdv
w 5.0. mm _ .0- 080.». m I 0 as 05 E 5000 00800 0:: 00:: 00:0 0000: m 0
0 m G U m <
08:00:00: 08:02.0:
05 05
E 82000 E 8280
80:25 80:25 8
8 2:008: 2:008:
3 80:00 80:00 .3 80:00 80:00 002880
.3 0280 .3 0280 0280 .3 0280 .3 0280 0280
£80 530 0 580 B 580 E 580 3 530 0
3 0:882 0:882 0:282 0:882 0:982 0:382
<0 <0 <0 <0 <0 <0
2:: 2:0
03:20 000< 0202-82 E 008:0 02:20 :02 0202 E 008:0
08:20 008
0208.8: 0:0 020—: :0 .0000 .8 0280 0:0 805% 02:80 .3 00:58:: <0 .8 800:: 0:0. 00:0 Ed 0300.

 

08:00.8:
05 E :2280
Emsoqw

8 2:00:83
3 80:00

E 8:00
5:00 .3

0208.8: 0:: 0085 :0 .0000 .8 0280 0:0 80:00. 322000 3 5:385 <0 .8 800:: 0:0.

80:00
0 SE:
580

.3 0:382

0280
580
.3 0:282

E :0280
08:25 8
2:00:00
3 80:00
3 0280
£80 3
0:582

80:00
E 8:0:
580 B
0:282

0028.80

0280
500C 3
0:382

 

08:20 0000
00:0 20 0300.

91

:02 .: 8:202:20 5.... 80m 0: 8:02:20 1. £002 20 8:002:20 ...

000052880 :2 00:00:20 2 .8 0:0; 00:80:? ”0082
~00N m2<0<2 0:: 0002 0:2 ”02:00

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_ mm" Z 082098000
mud mad NWO 0m .0 00. 0 00.0 008: m0-yl
Duo 000.0 02.2 * 20.2 020.0 m 2.2 34282 :0 2002-n!
00 2 00 My 00 2 00% 00 Mr 00 0, 002EE:Q 2m: 80:00-
$0.20 03.: ANN: £0.00 £0.00 €0.00
00m.0 n0m.0 00m. 0 ~30 05.0 03.0 0:20:00
000.: 30.00 30. 00 Q0. 0 80.0 $0.:
020.0- 000.0- 000.0- :0. - ...0—0.0- ...0N00- 0:0: 20:82: 02 00::2020
$2de $0.0 60.8 2.»... 0 Rm. 2 20 2w: 0
***Nm2.o- 030.003.0- 031.?“ .o- 03.30020. - tax—mod- ***mNc.O- 00-0w :2 00832230 :08 8202
:00 000.: 00: cm: 80.: 22.8
m00.0- 80.0- 80.0- 0000.0- ...»000- $000.0- AEEV 2.2 >02 - 80 E 0m:::0
80.00 200. 0 2:. 0 3.0.00 00.00 2 .00
000.0 000.: 000.: 80.0- 08.0- 80:- 00:6: 0: 0o 0.8» 5:880
80.00 30.00 80.00 0000 3.00: 2.000
mm0.0 £00 0w0.0 $0.0 20.0 000.0 00.2: 08020 8 0:
0 m D 0 m <
08:00:0:
08:88: E 82:00:
05 :2 8202000 00:25 8
00:25 02:00:02.. 0082880
8 02:00:02.2 80:00 3 80:00 80:00
3 80:00 3 0280 0280 3 0280 3 0280 0280
3 0280 5:00 5:0Q 5:00 .3 5:00 3 5:00 3
5:00 3 3 0:282 3 0:282 0:282 0:282 0:282
0:282 <01 <0 <0 <0 <0 <0

 

 

 

 

 

 

20.: 02:20 :0:< 002:2- :0Z :2 03::0

 

20.: 008420 :0:< 02:2 :2 0 :::0

 

 

 

0:: 020:. :0 5:00 .8 0280 0:: 805% 522000 3 32038.: <0 .8 800:: 0:0.

08:20 :00: 0:22:20:

02:0 Ed 03:0.

92

response to PA death or death of PA men, afflicted households reduce the non-maize
crop area first, while maintaining maize area in the short-term, later reducing maize area
in the medium term. The worst negative impacts are observed in the short-term when a
male head or spouse dies. Total area under cultivation declines by 1.2 hectares in the O —
3 years after such a death, with more maize area reduction felt than non-maize area
(0.81ha vs 0.38ha). In the medium term of 4 — 6 years after the death of a male head or
spouse, maize area declines by 0.53, a possible sign of recovery from death shock.
Beyond seven years after the PA death, no long-term effect is registered in change in area

planted.

3.6.7 Effects of PA Mortality on Agricultural Production through Change in Area

Planted among Afflicted Malawian Agricultural Households

Economic estimation results have so far showed that PA mortality leads to a
decline in area planted. Using area planted as a proxy measure for production we can
deduce that agricultural production will decline for afflicted households. Here, we
attempt to translate the estimated effects of PA death on area planted into an order-of-
magnitude estimate of reduced production by assuming common technology and input
use and average yields.

During field data collection, maize crop yields, were measured using agronomic
method of sampling yield of 5 x 5 meter subplots. The original plan was to assess hybrid
maize and non-hybrid maize separately to be able to compare effects across improved

hybrid and non-improved maize technologies, but due to logistical challenges it was not

93

possible to separate the two. In most cases, by the time enumerators got to into the field
for post harvest visit, farmers’ had mixed the harvests, hence our assuming common
technology use and average yields. Table 3.15 below shows the average per hectare
yield computed from these yield subplots for Blantyre, Kasungu and Mzuzu ADDS as

well as the average for the three.

Table 3.15 Average maize yields in kilograms per hectare estimated from yield
sub plots in Blantyre, Kasungu and Mzuzu ADDS

 

 

 

 

 

 

Region 2002 YSP Yield in Kg/ha
Blantyre ADD 1487
Mzuzu ADD 1566
Kasungu ADD 2247
Average 1752

 

 

Using the average yield figure of 1752 kg/ha, the death of a male head or spouse
estimated to cause a 0.35 ha reduction in maize area planted will result in average decline
in maize production of 0.6 tons per afflicted household. Maize constitutes more than
two-thirds of caloric consumption of Malawians (Gilbert et al, 2002). According to Smale
and Heisey (1997) Malawi has the largest per capita consumption of maize in the world
of over 150 kilograms per year. An afflicted household with an average family size of
5.2 would have a total maize consumption of 780 kilogram per year. The death of a male
head or spouse would mean such a household would fail to production 77% of its annual
maize intake. With low cash because hospital bills and funeral expenses, the household
also looses its ability to purchase food and thus become food insecurity. In terms of
effect of timing since death, reduction in production is highest in 0—3 years after death of

a male head or spouse followed by the 4-6 year interval- In this case the 0.81 ha and 0.53

94

 

ha reduction in maize area planted translate to 1.4 and 0.9 tons per year respectively
implying larger deficits in caloric intake among afflicted households.

Given the order—of—magnitude estimate results above we can reject the hypothesis
that there are no short-term, medium—term and long-term effects of PA mortality on
agricultural production. Agricultural production of maize does suffer a decline in the
short-term and medium term. The decline in the medium term is less than in the short-
term, showing a sign of recovery. No decline was observed in the long-term period of 7

to 13 years.

3.7 Conclusion

This study has shown first, that prime age adult mortality in Malawi reduces
agricultural production through a loss of area planted. Households with a PA death lost
between a fifth and a third of a hectare of area planted per each PA adult who died.
Second, non—maize area is more sensitive to prime age death shock than maize area
where a non—head or spouse death occurs. It appears afflicted households reduce area
planted to non-maize crops first while trying to sustain maize cultivated area to maintain
household food security. Non-maize crops also tend to require more purchased inputs
and labor than maize crop. Third, the gender and household position of the deceased are
important factors in distinguishing the magnitude of effect on area planted. The death of
prime age male household heads or spouses was more devastating in its effect to maize
area planted in both short and medium term, and non-maize area in the short-term. The
death of female household heads or spouses was not significantly impacting cropping

area. Fourth, monetary loss and farming experience loss associated with deceased PA

95

individuals are significant determinants of the magnitude of impact of a PA death shock
on total area planted. Fifth, the PA death effects are felt in the short — medium term (0 -6
years after death) and afflicted households respond by decreasing area planted to high
capital and labor input requiring non-maize crops in the short term and then reducing

maize area in the medium term.

Recommendations from the findings

We pose and answer four questions in this section on policy recommendations.
The questions are: Who is affected? What is affected? When is best time to intervene?
And how should intervention be made? We respond to each of these questions

separately below.

Who is affected?

Reduction in area planted and production among afflicted households is a major
concern for policy makers. As the results show, households who suffer loss of PA head
or spouse suffers the most in terms of loss in total area planted. This raises a question of
vulnerability of surviving widows. We recommend that policy makers recognize and use
the finding that gender and position in household of the deceased matters and use that in

whom to target for intervention.

96

What is affected?

Results show two distinct groups related to: a) the death of male heads or spouses
and b) all other significant PA deaths. The death of PA men who are heads or spouses
(group a) has the most devastating impact in terms of short—term decline in total area
planted, maize and non-maize area planted as well as medium term maize area planted.
Household suffering this shock will require interventions that target both maize and non-
maize crops in both the short and medium term. The second group shows non-maize
area more sensitive in the short-term and maize area in the medium term. These
households in general will have suffered the death of male others. For these households,
both maize and non-maize areas are affected but at different times thus requiring a

different response.

When is the best time to respond?
Households who suffer the death of male PA head or spouse will require
interventions in both the short and medium term while other households will only require

intervention in the short-term.

How should interventions be made?

Households suffering the death of a male PA head or spouse become widow
headed and are most vulnerable. We recommend direct food aid as well as targeted
inputs be given to minimize effect on food security. Government of Malawi in

conjunction with local NGOs can identify these households and offer assistance on a

97

sliding scale for the first 0-5 years. Such intervention will cushion the afflicted household
from food insecurity.

Households that lose non-head or spouse male adults or others short be targeted
for interventions primarily in the short-term. Our results suggest that such households
strive to maintain maize (food security) staple crop in the short-term by switching
resource from non-maize crop to maize crop. Providing these households with short-term
subsidized credit of seed, fertilizers and cash for labor-hire should enable then to
maintain non-cash cropping area in the short-term. These households can then use
earning from cash crops to mitigate observed medium term decline in the area under the

staple (food security) crop.
Areas of further research

Further research is needed to assess the effect of knowledge loss on agricultural
production as measured by other variables such as change in gross value of output. There
is also a need to assess to what extent the afflicted households recover their level of
production. Our research has shown that decline in area under maize in the medium term
is less than that of the first term among afflicted households, implying some kind of
recovery. More research is needed to knowing whether and how these households
recovery by tracking afflicted households annually to obtain more data points could help

further guide mitigation strategies.

98

A third area for further research pertains to the role of the elderly group, 60-79
years old. Further research is needed to find out more about their level of involvement in
mentoring prime age individuals so we can design programs to facilitate mentoring
between this group and pre—prime age groups to minimize. loss of experiential knowledge

in the rural setting.

99

References

Abbot, J ., M. Lenka, P. J. Lerotholi, M. Mahao, and S. Mokhamaleli. (2005). “From
condoms to cabbages: Rethinking agricultural interventions to mitigate the
impacts of HIV/AIDS in Lesotho”. A Paper presented at the lntemational
Conference on HIV/AIDS and Food and Nutrition Security, Durban, 14-16 April
2005.

Acton, J. (1975). “Non monetary factors in the demand for medical service: Some
empirical evidence’, Journal of Political Economy, 83.3: 595 — 614.

Ainsworth, M., G. Koda, G. Lwihula, P. Mujinja, M. Over and I. Semai. (1992).
“Measuring the Impact of Fatal Illness in Sub—Saharan Africa.” Living Standards
Measurement Study no. 90, World Bank, Washington, DC. 1992.

Ainsworth, M., and J. Dayton. (2003). “The Impact of AIDS Epidemic on the Health of
Other Persons in North Western Tanzania. World Development 31, (1): 131 — 48.

Ainsworth, M., and I. Semali. (1998). Who is most likely to die of AIDS? Socio-
economic Correlates of Adult Deaths in Kagera Region, Tanzania”. In
Confronting AIDS: Evidence from the Developing World, (ed). M. Ainsworth, L.
Fransen, and M. Over 95-110. Washington, DC: World Bank.

Ainsworth, M. and I. Semali. (2000). “The Impact of Adult Deaths on Children’s Health
in Northwestern Tanzania”. Paper from “The Economic Impact of Fatal adult

illness due to AIDS and other causes in Sub-Saharan Africa” Research Project.
World Bank, USAID and Danida.

Alderman, H.A., J .R. Behrman, H. Kohler, J .A. Maluccio, and SC. Watkins. (2000)
“Attrition in Longitudinal Household Survey Data: Some Tests for Three
Developing Country Samples”. Discussion Paper 96. Food Consumption and
Nutrition division of the lntemational Food Policy Research Institute.
Washington, DC. http://www.ifpri.org/divs/fcnd/dp/papers/fcndp96.pdf

Alderman, H.., J. Behrman, H. Kohler, J. Maluccio, and S. Watkins. (200]). “Attrition in
longitudinal household survey data: Some tests for three developing country
samples.” Demographic Research, 5278-124.

Alumira, J .D. (2002). “The Role of Technology in Agricultural Transformation in Africa:
Adoption and Impacts of Improved Cereal Varieties in Zimbabwe”. A paper
presented at the Workshop on the Green Revolution in Asia and its Transferability
to Africa. Tokyo, Japan. 8-10 December 2002.

Aventin L, and P. Huard. (2000). “The Cost of AIDS to three Manufacturing Firms in
Cote d’Ivore.” Journal of African Economies. 9 :161-188.

100

Barnett. T. (1994) “The effects of HIV/AIDS on Farming Systems and Rural Livelihoods
in Uganda, Tanzania and Zambia.” Final report to Food and Agricultural
Organization. Norwich, UK: Overseas Development Group, University of East
Anglia.

Barnett, T. and P. Blaikie. (1992). Aids in Africa: Its Present and Future Impacts. New
York: Guilford Press.

Barnett, T. J. Tumushabe, G. Bantebya, R. Sebuliba, J. Ngasongwa, D. Kapinga, M,
Ndelike, M. Drinwater, G. Mitti and M, Haslwimmer. (1995). “The Social and
Economic Impact of HIV/AIDS on Farming Systems and Livelihoods in Rural
Africa: Some Experience and Lessons from Uganda, Tanzania and Zambia. Field
Report”. Journal of lntemational Development. 72163-176.

Barnett, T., J. and A. Whiteside. (2002). AIDS in the Twenty-First Century: Disease and
Globalization. Palgrave Macmillan, Basingstoke, UK.

Barnum H.H, and L. Squire. (1979). “A Model of An Agricultural Household: Theory
and Evidence”. World Bank Occasional Staff Papers, No. 27. The World Bank.

Beegle, K. (1997) “The Impact of Adult Mortality in Agricultural Households: Evidence
from Rural Tanzania” PhD Dissertation, Department of Economics, Michigan
State University, East Lansing, Michigan.

Beegle, K., (2003), “Labor Effects of Adult Mortality in Tanzanian Households”, World
Bank Policy Research Working Paper no. 3062, World Bank, Washington, DC.

Bollinger, L. and J. Stover, (1999). ”The Economic Impact of AIDS”. The Futures Group
lntemational, Glastonbury, CT. USA.

Chapoto, A. (2006). The Impact of AIDS-related Prime-Age Mortality on Rural Farm
Households: Panel Survey Evidence from Zambia. PhD Dissertation. Department
of Agricultural Economics, Michigan State University, East Lansing, MI. USA

Chapoto, A. and TS. Jayne, (2005). Impact 0fHIV/AIDS —- Related Mortality on Rural
Farm Households in Zambia: Implications for Poverty Reduction Strategies.
Paper presented at IUSSP Seminar on Interaction between Poverty and
HIV/AIDS. 12-13 December 2005, Cape town, South Africa.

Chembezi, D.M. and A.W. Womack, (1991). “Program Participation and Acreage

Response Functions for US. Corn: A Regional Econometric Analysis”. Review of
Agricultural Economics. 13(2): 259-275.

101

Clay, E., C. Bohn, E.B. Blanco de Armas, S. Kabambe and H. Tchale. (2003) Malawi
and Southern Africa: Climatic Variability and Economic Performance. Working
Paper Series No. 7, Disaster Management Facility. The World Bank,
Washington, DC. pg 34—36,81

Cuddington, J .T. ( 1993) “Modelling the Macroeconomic Effects of AIDS, with an
Application to Tanzania”, World Bank Economic Review, 7: 172-189.

Cuddington J .T. and J. Hancock . (1994). “Assessing the Impact of AIDS: on the Growth
Path of the Malawian Economy”. Journal of Economic Development 43: 363-368.

Drinkwater, M., M. McEwan and F. Samuels. (2006). “The Effect of HIV/AIDS on
Agricultural Production Systems in Zambia: A Restudy 1992-2005. Analytical
Report. www.ifpri.org/reneanpdf/Zambia_AR.pdf

Donovan, C., and L. Bailey. (2005). “Understanding Rwandan agricultural household’
strategies to deal with prime age illness and death: A propensity score matching
approach.” A Paper presented at the lntemational Conference on HIV/AIDS and
Food and Nutrition Security, Durban, 14-16 April 2005.

Doward, A., and I. Mwale. (2005). “Labor market and wage impacts of HIV/AIDS in
rural Malawi.” A Paper presented at the lntemational Conference on HIV/AIDS
and Food and Nutrition Security, Durban, 14-16 April 2005.

Forgy, Larry. (1993). “Mitigating AIDS: The Economic Impacts of AIDS in Zambia and
Measures to Counter Them”. REDSO/ESO, February 1993. .

Forsythe, S., M. Roberts, J. Wangombe and O. Gakuru (1994) “Measuring the Impact of
HIV/AIDS on Africa’s Commercial Sector”. In Tenth lntemational Conference
on AIDS, Yokohama, Japan, 026D, p. 13. 1994.

Gavian, S., D. Galaty and G. Kombe (2005). “Multi—Sectoral HIV/AIDS Approaches in
Africa: How are they Evolving?” A Paper presented at the lntemational
Conference on HIV /AIDS and Food and Nutrition Security, Durban, 14-16 April
2005.

Gilbert, Robert A., Webster D. Sakala and Todd D. Benson. (2002) "Gender Analysis of
a Nationwide Cropping System Trial Survey in Malawi." African Studies
Quarterly 6, no.1: [online] URL: httpzl/webatfricanfl.edu/asq/v6/v6ila9.htm

Gillespie, S. (1989). “Potential Impact of AIDS on farming systems” Land Use Policy 4:
301-312

Houck, J .P., and ME. Ryan (1972). “Supply Analysis for Corn in the United States: The

Impact of Changing Government Programs”. American Journal of Agricultural
Economics. 54 (2): 184-181.

102

Jayne, T. S., M. Villarreal, P. Pingali, and G. Hemrich. (2005). “HIV/AIDS and the
agricultural sector in Eastern and Southern Africa: Anticipated Consequences.” A
Paper presented at the International Conference on HIV /AIDS and Food and
Nutrition Security, Durban, 14-16 April 2005.

Lee, DR, and P.G. Helmberger. (1985). “Estimating Supply Response in the Presence of
Farm Programs.” American Journal of Agricultural Economics. 67: 193-203.

Kambou, G., S. Devarajan, and M. Over, (1992). “The Economic Impact of AIDS in an
African Country with a Computable General Equilibrium Model of Cameroon”.
Journal of African Economies. 1: 109-130.

Kahn, K., S. Tollman, M. Garenne, and J. Gear (1999). “Who Dies of What?
Determining Cause of Death in South Africa’s Rural North-east.” Tropical
Medicine and International Health, 4, 433-441.

Mather, D., C. Donovan, T. Jayne and M. Weber. (2005). “Using Empirical Information
in the Era of HIV/AIDS to Inform Mitigation and Rural Development Strategies:

Selected Results from African Country Studies”. American Journal of
Agricultural Economics 87 (5): 1289-1297.

Morzuch, B.J., R.D. Weaver, and P.G. Helmberger (1980). “Wheat Acreage Supply
Response Under Changing Government Programs.” American Journal of
Agricultural Economics. 62: 29-37.

.Ndilu, M., D. Sequeira, and S. Hassigl. (1998). “Medical, Social and Economic Impact of
HIV Infection in a Large African Factory”, in IV lntemational Conference on
AIDS, Stockholm, Sweden, abstract 9583.

Over, M. (1992). “The Macroeconomic Impact of AIDS in Sub-Saharan Africa.” African
Technical Department, Technical Working Paper No. 3, World Bank,
Washington, DC.

Rugelema, G. (1998). “AIDS and Rural livelihoods: A case study of smallholder
households of Bukoba District, Tanzania”. PhD Thesis. Institute of Social Studies,
The Hague.

Rugalema, G. with S. Weigang and J. Mbwika. (1999). “HIV/AIDS and the commercial
agricultural sector of Kenya — impact, vulnerability, susceptibility and coping

strategies”. FAO/UNDP. 1999.

Sadoulet E. and A. de Janvry. (1995). Quantitative Development Policy Analysis. Johns
Hopkins U. Press. Baltimore, USA. 1995.

103

Smale, M., and P. W. Heisey. 1997. “Maize technology and productivity in Malawi”. pp.
63-79. In: Byerlee and Eicher, (eds.) Africa ’s Emerging Maize Revolution. Lynne
Reiner. Boulder, CO. 301 pp.

Smale, M., with Z.H.W. Kaunda, H.L Makina, MM.MK. Mkandawire, M.N.S. Msowoya,
D.J.E.K Mwale, and P.W. Heisey. (1991). “Chimanga Cha Makolo, Hybrids and

Composites: An Analysis of Farmers’ Adoption of Maize Technology in Malawi;
1989-91”. CIMMYT Economics Working Paper 91/04. Mexico, D.F.: CIMMYT.

Strauss J. and D. Thomas. (1995). Human resources: empirical modeling of household
and family decisions, in TN. Srivasan and J. Behrman (eds), Handbook of
Development Economics, Volume 3, Amsterdam: North Holland Press,
Introduction.

Tibaijuka, AK. (1997) “AIDS and Economic Welfare in Peasant Agriculture: Case
Studies from Kagabiro Village, Kagera Region, Tanzania”. World Development;
25: 963-975.

UNAIDS. (2005). “Report on the Global HIV/AIDS Epidemic”. UNAIDS, June.

UNAIDS. (2006) “Report on the Global HIV/AIDS Epidemic 2006”. UNAIDS, Geneva,
Switzerland. http://www.unaids.org[en/HIV%5Fdata/2006GlobalRemrt/

UNAIDS/W HO. (2005). “Malawi: Epidemiological Fact Sheet on HIV/AIDS and
sexually transmitted diseases”. UNAIDS/WHO Working Group on Global
HIV/AIDS and STD Surveillance. Geneva, Switzerland. _
Http://who.ch/emc/diseases/hiv. , http://www.unaids.org and
http://www.aids.africa.com. 2005.

Urass, M., J .T. Boerma, R. Isingo, J. Ngalula, J. Ng’weshemi, G. Mwaluko, and B. Zaba,
(2001). “The Impact of HIV/AIDS on Mortality on Mortality and Household
Mobility in Rural Tanzania”. AIDS, 15, 2017-2023.

US Census Bureau, (2002). The AIDS Pandemic in the 21" Century. Report for XIV
lntemational Conference on AIDS, Barcelona, July 2002, lntemational Program
Center, US Census Bureau.

Yamano, Takashi, and T. S. Jayne. (2004). “Measuring the Impacts of Working-Age
Adult Mortality on Small—Scale Farm Households in Kenya.” World
Development, 32 (1): 91—119.

Yamano, T and TS. Jayne. (2005). “Working-age Adult Mortality and Primary School

Attendance in Rural Kenya.” Economic Development and Cultural Changes. Vol
53 (3): 620—621

104

Yang, Dennis T., and Mark Yuying An, (2002). “Human Capital, Entrepreneurship, and
Farm Household Earning,” Journal of Development Economics, 68: 65-88

Wooldridge, J. M. (2000). Introductory Econometric: A Modern Approach. South-
Westem College Publishing. USA.

World Bank. (2000). Intensifiring Action against HI V/AIDS in Africa: Responding to a
Development Crisis. Africa Region. The World Bank. Washington DC.

105

CHAPTER 4
NON-FARM LABOR ALLOCATION DECISIONS AMONG MALAWI’S RURAL
SMALLHOLDER HOUSEHOLDS IN AN ERA OF HIV/AIDS
4.0 Introduction

Rural non-farm employment (RNFE) has been argued to be a significant
contributor to rural household economies through diversifying sources of livelihoods in
order to reduce income risk, maintain or improve food security (income and
consumption) and generate cash for other cash consumption goods (Ellis and Allison,
2004). Indeed, income from the non-farm sector has been reported to generate positive
effects on agriculture and complement the absence of credit markets. Evans and Ngau
(1991) observed that non-farm income generates cash that can be used for purchasing
farm inputs and hiring labor to improve farm yields.

While non-farm income from RNFE could potentially benefit households that
participate, the critical question becomes this: Do the households have the capacity and
willingness to participate and tap into those potential benefits? For a household to
participate it has to have not only prime-age adult members who are capable of
participating (capacity) but also ones who choose to participate in rural non-farm
employment (willingness). Prime agel adult death in the era of HIV/AIDS reduces the
amount of these prime age adults who can potentially participate in rural non-farm
employment during long-illness period, death and funeral events as well as in the post
funeral period. A prime-age-death afflicted household making labor allocation decisions
between (own) agricultural and non-farm activities could, on the one hand potentially

choose to use all its labor on-farm and even hire supplemental labor and thus elect not to

 

' Prime-age adult (PA) is defined as individuals 15-59 years old

106

participate in the non-farm employment because of the labor constraint. On the other
hand, the household could elect to participate in rural non-farm employment to generate
additional cash lost during illness, death and funeral expenses. In this case the household
decides to send its members out because of the cash constraint.

The characteristics of the deceased matter as well. Whether the deceased was a
head of household or spouse or not or male or female could affect the decision power
base of the household. For instance, if the deceased was a head or spouse, decision-
making capacity will decline during illness and reach a low level when the individual is
incapacitated or dies. The household will have to regroup over time after the death
impact. Female adults who usually take nursing roles may be forced not to participate in
non-farm employment during illness and death but would need to go out in full force
after the death. On the contrary, if orphans survive the deceased, prime age women may
have to stay on farm to look after orphans and therefore not participate in rural non—farm
employment. These ambiguous household labor allocation decisions could be initial
responses to a death shock. As the shock wanes with time elapsing after the death shock,
the household could make decisions different from the initial response. These contrasting
effects therefore raise empirical questions that could be best answered by assessing the
effects of prime age death on off-farm employment participation.

Numerous studies have recommended promotion of non-farm income generating
activities or forms of livelihoods to help mitigate AIDS impact on afflicted households,
based on anecdotal evidence. This study provides empirical evidence as to whether or

not i) afflicted households are participating and if they are, what factors are determining

107

their intensity of participation, and ii) whether or not gender of the deceased and time

elapsed since death matters when it comes to off-farm labor allocation decisions.

4.1 Previous Research on Rural Non-Farm Employment and Adult death

Since the mid 1990’s perceptions of the rural non-farm sector have shifted from
being viewed as low-productivity, low-quality output sector to one viewed as an
important contributor to economic development and economic growth as claimed by
Reardon et al., (2001). Recognition of the importance of RN FE as an engine of economic
growth spurred a flurry of studies to identify the determinants and benefits of non-farm
activities and incomes in rural households in Latin America and Africa. These studies
produced evidence that RNFE provides important distributional effects in providing
employment to rural poor and landless, which helps achieve income smoothing through
diversification, strengthening food security, reducing poverty and curbing rural urban
migration (Lanjouw and Lanjouw, 2001). These studies went on to assess the
determinants of participation and intensity levels with a view to advising policy makers
on the constraints that act as barriers for expansion of RNFE. Three groups of
determining factors were identified (i) individual characteristics (such as age, gender,
education), ii) household/farm characteristics (such as number of adults, land area,
productive assets), iii) location-accessibility attributes (agro ecological regions, proximity
to urban areas, altitude, infrastructure), see (Ruben and Van Den Berg (2001), Ferreira
and Lanjouw (2001), Coral and Reardon (2001), Sanchez (2005) and Lanjouw et. al,
(2001). One thing very evident from the literature was that determinants tend to be

location specific. While common determinants were identified, location specificity of the

108

results dictates that a localized study should be conducted from which effective local,
regional and national intervention programs could be developed.

The above-mentioned studies where largely silent on the effect of PA death on
RNFE labor decisions. This is a critical issue in the era when HIV/AIDS is ravaging the
prime—age adults who form the labor pool with greatest potential to participate in RNFE.
Omission of PA death variables in these studies could result in biased coefficient
estimates if omitted death variables have significant effects. Below are five reasons that
death variables should be included in the non-farm labor allocation decision analysis.

First, death reduces the stock of labor available to participate in RNFE. Where
HIV/AIDS death is involved, labor stock is affected during the long illness during which
nursing care is needed. Medical and funeral expenses reduce the amount of money
available for farm household cash needs like hiring extra labor or purchasing other family
labor substituting inputs such as herbicides during the morbidity/mortality period. Huge
medical and funeral expenses could also affect the resiliency2 of the? household over time.
Death therefore potentially affects the RNFE participation decision at the time it occurs
as well as after death has occurred.

Second, the RNFE literature makes a distinction between the decision to
participate and the decision on the level of participation (intensity). In the context of
studying RNFE and prime-age adult mortality (PAM) this becomes very important. The
decision on whether or not to participate in the RNFE market may be socially or
psychologically driven where the individual may prefer not to work off-farm no matter

the payoff. For instance, a PA individual may decide to abstain from participation to

 

2 Resiliency is defined as ability to recover from an adverse shock or disruptive change without being
overwhelmed - http://www.resiliencycenter.com/definitions.shtml

109

mourn the loss of a loved one. That decision could also be taken because the enticements
fall below the minimum level that person would require participating in non-farm
employment (Matshe and Young, 2004). This participation decision could potentially be
influenced by prime-age adult death. The intensity of participation is the next decision
about how many hours to work, when the adult from an adult death afflicted household
has decided to participate. At this level too, the individual may decide to work fewer
hours so as to have more time at home in order to fill in the void left by the deceased.
Alternatively, the adult may work more hours to raise more money to recover from the
financial loss related to lost contributions of the deceased or the medical and funeral
expenses.

Third, research has indicated that there is a distinction between skilled non-farm
and non-skilled non-farm employment opportunities. De J anvry and Sadoulet (2001)
found the level of education to be a significant determinant of accessing better paying
skilled and semi-skilled non-agricultural employment, but not for agricultural non-farm
employment. But bringing adult death into the mix could change this. Adult death
affects the educational composition of the household. Earlier studies provided evidence
that higher HIV rates occurred among better educated/higher income groups (World
Bank, 1999), while more recent studies do not indicate such a correlation (De Walque,
2006). While it may be unclear who is dying from HIV/AIDS, what is clear is, that adult
deaths affect the educational composition of the household, which will then have an
effect on which types of non-farm jobs surviving members can take.

Fourth, the gender composition of a household is critical in Malawi, where

patrilineal and matrilineal societies exist, because power relations and access to resources

110

differ between these societies. Generally men have more dominance in patrilineal society
and women do in matrilineal societies. In terms of access to off-farm employment
opportunities, gender may play out differently where non—farm non-agricultural
businesses may be male dominated. In such circumstances the death of a prime age man
may rob a household of an adult who could participate in the non-farm employment
whereas a female death would not. Where the non-farm sector is more under female
domination, results may be to the contrary. Another factor about gender has to do with
division of farm household-labor. Chores that are predominantly labeled female work —
such as cooking, childcare and nursing the sick - may prejudice participation by female
prime age adult in off-farm employment. Therefore the gender of the deceased and the
gender mix among surviving household members could potentially affect participation in
non-farm employment.

Fifth, prime age death potentially induces a push into participation or a pull out of
participating in the non-farm employment. If the deceased was a strong labor
contributor either physically or by funding labor hiring, their labor input shortfall could
force an afflicted household to pull out of the non-farm sector and concentrate on their
own farm, creating a “labor pull” effect. On the other hand, if the death of an individual
prime age person leaves the family in a financial hole due to large expenditure and cash
disbursement, death may force the household to send surviving members out into the
non-farm employment to raise cash, creating a “financial push” effect.

These potential responses could also be initial reactions to the death shock that

could reverse over time when the household adjusts to a new equilibrium. There could

lll

therefore be an initial response and a post adjustment period response that could show
recovery and or resiliency of the household to the death shock.

Rural Non Farm Employment classification: The literature makes a distinction
between types of RNFE by separating agricultural wage. labor (unskilled) from non-wage
self-employment (semi-skilled) from salaried labor (skilled). Barrett, et al (2001)
disaggregated rural employment into wage employment and self-employment in
agricultural and nonagricultural sectors. Escobal (2001) disaggregated non-farm
employment, self-employment, and wage employment into high skilled and low skilled.
Dirven (2004) divided the job types into low and high productivity to emphasize the
different entry requirement. In his study, Dirven (2004) reported that low entry barriers
and low rates of return jobs characterized by low skilled activities, were deemed
accessible employment by job seekers with low qualification. In this study we
disaggregate rural non-farm employment into two categories, agricultural RNFE (ag-
RNFE) and non-agricultural RNFE (non-agRNFE) as shown in Table 4.1 below.

Table 4.1 Classification of types of rural non-farm employment activities from
rural Malawi

 

 

Classification Abbreviated Types of activities included
Notation

Agricultural rural non-farm agRNFE Casual ganyu labor,

employment dimba gardens,

Estate wage labor

 

 

Non agricultural rural non- non-agRNFE Beer brewing,

farm employment buying & selling (Geni),
contract work, fishing, capentry
selling fruits, selling firewood

 

 

 

Source: MAPAMS 2002
The agricultural RNFE requires low-level skills, and involves manual labor like

weeding and harvesting at the direction or supervision of owners. These jobs are located

112

 

with walking distance or same locality and have lower barriers to entry. The non-
agricultural RNFE, on the other hand, tends to involve traveling (with the exception of
beer brewing). Non-agricultural RNFE also requires some skill level - technical know
how (e. g., carpentry and contract work) and bargaining skills (trading ——gem). Most
important, non-agRNFE requires some capital outlay that could bar women (in patrilineal
societies), and poor households from participating. Non-agricultural rural non-farm

employment therefore represents a category with higher barriers to entry.

4.2 Study Objectives

A review of the literature on RNFE and adult death shows a disproportionately
large volume of papers that focus on RNFE excluding adult death. Very few studies
(Beegle, 2003; Anglewicz, 2005) look at RNFE and adult death. Exclusion of death
variables in RNFE literature could lead to biased coefficient estimates if death variables
influence RNFE labor allocation decisions in the era of HIV/AIDS. This is an important
knowledge gap for which more studies are needed. This essay attempts to fill this gap in
the literature by going beyond anecdotal evidence to empirically assess the relationship
between PA adult death and RNFE. The broad objective of this essay is therefore to
evaluate how death stricken households respond to RNFE labor markets and how their
response changes over an adjustment period following the initial death shock. The
specific research questions for the prime-age adult death stricken households are:

1) What are the determinants of an individual’s participation in agricultural and non-

agricultural rural non-farm employment in rural Malawi?

113

2) What are the determinants for the level of intensity in participation measured by
level of labor allocation in agRNFE and non-agRNFE?

3) Is there a difference between initial response and response after some period?
How long is the adjustment period and what canwe learn about household
resiliency over time?

4) Given empirical evidence of determinants of participation and intensity, what
policy measures (if any) need to be formulated to support participation in RNFE

in the era of HIV /AIDS?

This study uses data from 351 farm households in 2002 Malawian to provide
answers to the research questions on how prime-age adult mortality (PAM) stricken
households respond to rural non-farm employment labor market and how their responses
change over an adjustment period. The analysis employs a double hurdle econometric
model that examines separately the determinants of the decision onwhether to participate
and those of the decision on how much time spent participating in the rural non-farm

labor market.

4.3 Conceptual Model

In order to analyze the determinants of individual participation and level of
participation in RNFE, we use a rural household economic model to explore the labor

supply decision. We assume a household where each member i (i: male (m) or female
(f)) derives utility from consumption of leisure time (including home chores) T; and

consumption goods Y purchased from the market. The utility attained is also conditional

114

on individual characteristics (Zi ) like human capital and taste, household characteristics
(Zh ) such as family size, and the prime age death status of the household (Z d ). The

economic decision-making framework of these households is summarized in equations

below.

i
Max U(Tl 'lei’zh'zdl

Tgf,r}a,r;,r

Subject to:

T=T}a+r,ff+rf, (1)
Q: Q(Tia,X,-Zi,Zq,Zd) (2)
wilf 1;]. + PqQ(Tia,X;Zi,Zq,Zd )— PxX + Remit— PyY 2 0 (3)
T; 2 0; T21 20,75}. 2 0, (4)

where i = m, f; nf = agRNFE,non—agRNFE

The above framework depicts a rural farm household that receives human time

endowment each year ( T ) for prime-age adult members of the family. T}? is labor time

allocated among work on their own farm by each member i (i=m, f) , Thf is labor time

allocated to non-farm work disaggregated between agricultural rural non-farm labor time

(TagRNFE) and non agricultural rural non-farm labor time (Tmm_ agRNFE ), and T1 18

leisure time for each member. Equation 2 is the technology of farm production

represented by the concave production function where output Q depends on labor time

115

and other purchased inputs conditional on individual characteristics (2i ), farm

characteristics (Zq) and whether the household has had a prime-age adult death or not

(Z d ). Equation (3) depicts the budget or cash income constraint where money spent on

consumption goods (Y) at price ( Py) must not exceed the household income composed
of money earned from non-farm work through the sale of Thf labor time at a wage of

w; f , net farm income from sales of agricultural produce netting out money spent on
purchasing inputs including hired labor (PqQ- PxX ) , and other exogenous incomes such

as remittances (Remit). Equation (4) is the non-negativity constraint that ensures interior

solution for all labor allocation variables in optimization.

The off-farm wage (w;z f ) offered to prime-age individuals i (i=m, f) are assumed
to depend on their respective marketable human capital skills (Zt) such as education,

vocation training and experience and local'labor market characteristics (a) such as
distance to work place and commuting time. This off-farm wage rate as a function can be
stated as equation 5 below.

whf =w'(Zl-,(p) (5)

Location and poor rural infrastructure that relate to commuting distance and associated
travel expenses could potentially reduce labor mobility in rural labor markets and affect
wage offers.

The death status of a household (2 d ) can be formally stated as in equation 6

below.

116

zd = D(D_{(0.1).YSD_{),

Where j = m, f ; s = hos (head or spouse), nhos (non head or spouse) (6)
D is a vector of variables that identifies the prime age adult mortality status in the
household (2 d ). Dsj is a dummy variable for prime age adult death where D! = 1 indicates

a household stricken by a prime age death of an individual j ( j = m, f ) whose status (s)

or position in household is either head of household or spouse (has) or non head or
spouse (nhos). YSDSj is the number of years since the death of individual j of status 3

occurred in the household. This is a quantitative variable that measures the number of

years that elapsed since a prime age death occurred in a household. It is used to capture

post adjustment effects. When Dsj = 0, no prime age death was observed in the household

and YSDSj = 0. Off-farm labor decisions are affected following the funeral in that hours

allocated will either increase or decrease depending on whether the-household
experiences labor stress or financial stress.

The key household decision or choice variables in this study are

X ,T}a,T,€f ,7"; ,Y given 2 d . The conditions for optimal decisions are obtained by

maximizing the utility function subject to resource constraints imposed by the time
constraint (1) and the income constraint (3). The Lagrangean expression associated with

this utility maximization problem is:

L=U(Tli,Y)+/l(wflf7;€f +PqQ(.)— PxX+Remit—Pyl’)+r(T-T}a —T,';f 4,") (7)

The first order conditions are:

X: ,1”qu — Px 1 = 0 (8)

117

T}; ,1[PdQTJ.t—r']= 0 (9)

Awif(a))—z'i so,

T'f >0,T'f(/i.wf—1f 20:0 ' (10)
T,’.- UT;- —ri =0; (11)
y: uy—zpy =0 (12)
,1: wnf Tnf +P qQ(.—) P X+Remit-P r (13)
r: T— T' —T‘ —T1 =0,- (14)

f(1 nf

where i: m,f; nf = agRNFE,non —agRNFE

The symbols 2. and r are Lagrange multipliers for marginal utility of prime age

individuals’ income and time respectively, while UT i and U ,are partial derivatives of
l

the utility function subject to leisure time T; and consumption goods Y respectively.

Q i and Q x are partial derivatives of the production function with respect farm labor
T
fa

time and purchased inputs respectively.

The first order conditions above adopt a simplified model farm household with
two prime age adult individuals, a male and a female, following the Huffman and Lange
( 1989), and Tokle and Huffman (1991) theoretical framework for husband and wife
scenario. Equations (9) — (l 1) provide conditions for optimal time allocation by the two
individuals for which farm labor time and leisure time are assumed to be positive as (9)

and (11) are equalities. Equation (10) gives the optimality conditions for non-farm work.

118

If wi ( w)< 7% , then the optimal non-farm hours is zero (Trifle = 0 ), however, if

wi (w) = 11% then the individual’s non-farm wage, net of commuting cost, equals the
marginal value of his or her leisure time or farm labor time, and optimal non-farm work
may be positive.

When an interior solution for Tnf occurs, the non-farm wage ( w; f ( a) )) determines

the marginal value of the individuals’ time. Equations (8)-(10) are then the conditions for
profit maximizing farm input usage, and can be solved independently of the rest of the

equations to obtain the farm input demand functions including farm labor demands for

the individual (TJZ) and purchased farm inputs ( X *) as shown below:

T

f*,*x=F(wnf, w,,foPq,.,,ZZZd) (15)

W I (1’
To obtain demand functions D, for male and female prime age individuals’ leisure time,

equations (6), (11)—(l4), conditional on (8)-(10) are used to obtain equation 16 below.

' Wf
Tl” =0,“wa n,f,,n’Py Remit,Z. ,zh 2d);7r= Pq’k—Q —PX *—w,;fT}* (16)

The non-farm labor demand functions are derived using the total human time endowment

constraint (equation 1). In equilibrium demand for non-farm labor ( T; f )equals

supply( Sl ) such thatT — -S'* .When the both male and female PA individuals
f T'f lf

participate in the non-farm employment;

T‘* =T— T'*—T'*
nf fa

=S'f(w zf’ wifo, ’Pq’ P), ,Remit,Zl. "Zh ,,Zq ,Zd) ; i=m,I (17)

119

When the non-negativity constraint is binding for either the male or female PA, farm

production decisions are inseparable from household consumption and labor-supply

decisions. Though non-farm labor supply continues to be derived residually, Ti; and

T; *contains the same variables.

Since this study is focusing on the labor response of PA death afflicted household
to RNFE labor markets, then we substitute equations (6)-(7) into the labor supply
function (17) above to obtain the general labor supply function that captures death status

and location effects to be estimated in this essay:

Ti*-T—Ti*—Ti*

nf — fa l
_ m f - i j . -_
—Si(wnf,wnf,Px,Pq,Py,Remit,Zi,Zh,Zq,(p,Ds,YSDS ) , t—m,l (18)

The farm-household model does not offer determinate apriori predictions of the effects of
the change in the exogenous variables on equilibrium quantities of non-farm work. The
wage elasticity of supply of non-farm hours could be either negative, positive or zero. For
a participating individual for whom leisure is a normal good, an exogenous increase in
non—farm wage could have ambiguous effects on the quantity of non-farm hours worked.
A pure substitution effect, holding utility constant, decreases leisure time, but the income
effect increases demand for leisure time, thus having an opposite effect on non—farm
hours worked. For a non-participant, a rise in non-farm wage increases the probability of
becoming a participant. A death of a prime-age individual could reduce exogenous

income and wealth, forcing prime age individuals into the non-farm labor market to

120

recover lost income, a financial push effect, or contrarily, it could force the individual out
of the market to concentrate on farm labor or leisure (home time to nurture orphans).
These effects could be reversed as time elapses after the death shock. The expected
effects of gender, location, wage increase, prime age adult death on rural non-farm
employment are therefore generally ambiguous apriori and therefore should be

determined empirically.

4.3.1 Modeling Non-farm Work Decisions

While the utility is maximized at the household level, each prime age individual
decides whether to participate or not. If they decide to participate, they must also decide
how many hours to participate in the non-farm employment (Benjamin and Guyomard,
1994). Some jobs require skills, experience and or capital so these factors become
critical for these decisions. An individual’s decision to participate in non-farm

employment will depend on a comparison of the market rate and individual’s reservation

wage. Technically, the reservation wage, Wh} , for individual i (i = m, f) is the marginal
value of his/her labor when no time is allocated to non-farm work. Thus an individual

i (i = m, f) will only participate if the market wage, w: 1f , exceeds the reservation

wage, is)”; . Therefore, the probability of a PA adult participating in the non-farm

employment can be expressed as the probability that the reservation wage is lower than

the market wage leading to the following decision rule:

i_ - ,ri .i -_
Tnf—O’lfwnfzwnf 1—m.f

121

Trif>0,if w;}<wfy. i=m,f (19)

The reservation wage, w "f, is an intrinsic variable that depends on the vector of

exogenous factors V like input prices, variable output prices, exogenous household, farm,

individual, death status and local labor market characteristics such that

wf j _ j . .
,,f (V) W,',' f(w,’,_"f an ,,Px P, qHPy ,Remit,Z. ,,zh zq,¢,os,ysos ). The varlables that rarse

the reservation wage reduce the probability of non-farm work, but the variables that raise
the non-farm wage rate increase this probability.

The model for estimating non-farm participation can then be written as:
w; = V'P. + 81. i: f,m, nf = agRNFE,non —agRNFE (20)
where ,6,- are estimated coefficients, and e,- is the random disturbance.

If we define a binary variable 9’ that equals one if member iparticipates in the rural
non-farm employment, then the RNFE participation decision rule can be expressed in

probabilistic terms as stated in equation (21).

Pr[ Téf > 0] = Pr/oi = 1] i: f,m, nf = agRNFE,non -agRNFE

=Prlwnf<whf]’

=Pr[V’,B..+£ <wnf]

=Pr[e. <wf _v'P,1 (21)

In equation (21), the probability of observing hours worked is equated with the

probability of Qi equal to one or the probability of observing participation.

122

The non-farm employment literature makes a distinction between the decisions (1)
whether or not to participate in the RNFE labor market and; (ii) the amount of time the
participant allocates to non-farm work (Matshe and Young, 2004). Some studies have
focused solely on participation decision (Lanjouw, 2001 and Ruben and Van den Berg,
2001) and ignored the potentially important “hours worked” or participation intensity
decision. Other studies have modeled these as decisions made jointly where analysts
focused only on the hours worked arguing that if there are hours worked recorded, it is
obvious the individual decided to work as shown in equation (21). Matshe and Young
(2004) point out a critical reason why joint-decision making could lead to erroneous
conclusions. A zero observation on hours worked could be a result of personal
preferences, inadequate qualifications or disability, or inadequate economic incentives. It
is therefore critical to separate those who are incapable of participating, from those who
chose to abstain from participating no matter what level of exogenous variables
(abstention), and from those who only participate for certain levels’of relevant variables

(comer solutions). We discuss joint versus separate decision making next.

4.3.2 Modeling Joint Versus Separate Participation and Intensity Decisions

In this essay, we seek to understand both the factors affecting the decision to
participate in non-farm work and those affecting the decision on intensity levels (hours
worked). These decisions can be made jointly or separately as mentioned above. When
decisions are made jointly, Tobit models can be used to infer what influences the

decisions made. This has been the traditional way of analyzing joint decisions (Greene,

123

2002). Unfortunately for non-farm studies, it implies that individuals for whom zero work
hours are observed are labor market participants who choose not to work at current level
of exogenous factors like economic incentives (going wages). It rules out the possibility
that a zero observation may not be a comer solution, but the fact that the individual may
abstain from participating due to inadequate qualifications, disability or personal
preferences. Further, Tobit regression restricts the determinants of participation and level
of intensity to the same variables, which in real life may not be realistic.

Participation and intensity level decisions are not necessarily made jointly. The
decision to participate could precede the decision on the intensity level of participation.
The double hurdle model, first developed by Cragg (1971), offers an appropriate general
approach to modeling participation and hours worked as two separate sequential
decisions. At the first stage, an individual decides to participate in the labor market, and
at the second level, the individual allocates some time to work non-farm. This is

modeled by assuming the existence of two latent variables: 9',“ associated with the

decision to participate in the RNFE labor market (same as (II in Equation 21) and 91.2“ ,

linked to the decision on how many hours to work off farm. The variables are a linear

function of the first and second vector of hurdle regressors, VI and V2 respectively.

9'1“ :Vl'6i1+#il ’ (22)

i**_

Q2 ' 2312 “‘12, (23)
For equation (22) and (23) V1 represents the participation decision explanatory variables

and V2 represents those that explain the decision on hours of work or intensity levels of

participation by individual i, (i = f ,m ).

124

Participation Decision

If we let 91* denote an index variable that equals one if individual i decides to

participate, and equal zero otherwise, then:

i**

o‘,*=1,if n] >0,
and

(2‘: = 0 , otherwise. (24)

Equation (24) represents the first hurdle. If we assume the error term fin in Eqn (22) to be

normally distributed, the first hurdle can be modeled as a probit.

Intensity Level Decision

The intensity level decision corresponds to choosing the hours to work in the non-
farm employment (2’ * , conditional on clearing the first hurdle. The RNFE labor
allocation can be generated as below:

o‘*—oi**,if oi**>0,

2 _ 2 1
and
Q"; = 0, otherwise, (25)

125

Equation (25) represents the second hurdle that can have zero level of non-farm labor
independent of the first hurdle and it takes the form of a Tobit model. To get the actual

hours observed of work, T5; = Qi , we interact both hurdles such that:

o = (2', ”'9‘; (26)
The interaction in equation (26) allows zero work hours to be generated by a failure to
clear either or both of the hurdles. When zero level of non-farm labor are observed
(among positive hours) independent of the first hurdle, this implies joint decision-
making, which takes the form of a Tobit model. The Tobit model is applied to outcome
variables that are roughly continuous over positive variables but have a positive
probability of equaling zero (W ooldridge, 2000; p551). However, when we observe non-
zero positive hours of work in the rural non-farm employment, this implies the individual
must be both a participant in the market and has decided on a positive level of work time.
In this case he or she made two separate decisions sequentially, clearing two hurdles —
hence a double hurdle approach.

We assume the latent variables have a bivariate normal distribution such that
( p“. 412i ) ~ BVN(0,2 ), where 2: =[ 1 po] (27)

.00" 02

The bivariate normal distribution implies that the Tobit model is nested within the double
hurdle model making the Tobit model a restricted model and the Cragg model an
unrestricted model (Greene, 2002). When probability of participation is assumed to be

equal to one (p=0), Equation 27 becomes the more parsimonious Tobit model.

126

When do we apply double hurdle versus Tobit models? Are decisions on whether
to participate and how many hours to work being made jointly or separately? By
choosing one model over another, a researcher makes an implicit judgement about
whether decisions are made jointly or separately. Lanjouw, 2001 and Ferreira and
Lanjouw, 2001 used the probit model to estimate participation, ignoring the potentially
important intensity decision. Escobal (2001) used a Tobit double-censored estimation that
focused on income shares (an intensity measure), ignoring the participation decision.
Matshe and Young (2004) found out that participation and hours worked decisions are
not based on the same decision making process. They observed that modeling the
participation decisions separately from participation intensity level decisions separately
seems to be more appropriate and realistic even though Tobit may be more parsimonious.
Whether a Tobit or double hurdle model is more appropriate can be determined by
running the two models separately and conducting a likelihood ratio test that compares
the Tobit with the sum of likelihood functions of the probit and truncated regression
models (Greene, 2002).

Greene (2002) suggested that these two models could be compared using a
likelihood test ratio that takes the double hurdle model to be the unrestricted model and

the Tobit model the restricted model. The likelihood ratio statistic, LR, can be computed
as LR =-2[ln i R - ln LU ], where 1n 1: R is the log-likelihood function of the restricted
Tobit model estimation and ln I. R is the log-likelihood function from the probit model and

compared with chi-square table result to test the null hypothesis that p = 0 , that is, Tobit

model is the preferred mode]. We will use this for choosing the underlying decision

process in this essay.

127

4.4 Hypotheses

Based on the conceptual framework above, five hypotheses can be postulated for
empirical testing. The first hypothesis relates to selecting the underlying decision process,
i.e., whether decisions are made jointly or separately. The remaining hypotheses are
divided between factors that affect individual decisions to participate in rural non-farm
employment and those factors that affect the degree of participation or intensity levels in
term of hours worked.

There are four key underlying factors that buttress the hypotheses on participation
and level of intensity of participation for this study. 1) A household head or spouse has
leadership responsibility on farm household; the death of such a PA individual leave a
gap in leadership that is likely to draw prime age adults back on farm to fill this
leadership void. 2) Non-head or spouse members are free from leadership responsibility.
They can therefore participate more in the non-farm employment sector as money-
spinners for the household. The death of these PA individuals would create a financial
loss that will more likely motivate PA individuals to enter the market to try and meet the
financial gap — a “financial push” response. 3) A gender barrier exists in rural Malawi
where lucrative non-farm employment eludes the women. The death of a PA woman
would not cause movements of individuals into non-agricultural RNFE because the
deceased would not have been part of that market in the first place. Rather we expect a
muted response toward the ag__RNFE, though it has lower barriers to entry for women,
due to the home care labor burden, which is predominantly perceived to be women’s
responsibility. The non-agricultural Rural Non Farm Employment is assumed to be a high

barrier to entry market with better pay offs and generally skewed in favor of men

128

(Hussein and Nelson, 1998). 4) The rural non-farm employment market in Malawi is
itself a thin market that limits opportunities anyway. RNFE may not support household
resiliency by allowing adequate growth of individual participation and intensity levels

over time.

4.4.1 Decision Making Process Hypothesis

HP, The null hypothesis is that contrary to results by Matshe and Young
(2004), the participation and intensity decisions are made jointly. This implies the
parsimonious Tobit model will be the preferable model of choice versus the double

hurdle model.

4.4.2 Participation Hypotheses

HP; The death of a male head of household or spouse (hos) could reduce both
finances and household stock of labor time. We therefore expect surviving PA adults to
shun participating in agRNFE in order to fill the labor void left by the deceased. The
incentives from the low remunerations of the low entry barrier agRNFE market won’t
exceed the opportunity cost, or reservation wage for participation to occur. If the
deceased PA individual was a female head or spouse, it is likely that the deceased was
involved in the agRNFE bringing in some additional income. However, their labor
contribution gap will be felt mostly on-farm if they provided childcare. Surviving PA

adults would have to stay on farm to meet that labor demand. We therefore hypothesize

129

that the likelihood of adult participation in agRNFE would decline due to a strong labor

pull in a household afflicted by the death of either a male or female head of household or
spouse. This implies that, pr(agRNFEp|DSj = 1 )< 0 , for j: m, f and s: has , where

agRNFEp represents participation in agRNFE.

HP3 The death of a PA man, regardless of his position in the household will
influence adults to participate in the RNFE, both agricultural and non-agricultural, while
the death of a woman irrespective of her position in the household will influence
individuals to pull out of the non-farm labor market. This hypothesis essentially declares
that a man’s death is associated with a financial loss which gives PA individuals a
“financial push” to try and recover the lost finances, while a woman’s death is associated
with a loss of labor on-farm and on the home front, hence PA individuals will refrain
from participation, to make up for lost labor - a “labor pull” effect.‘It makes no

distinction between household status and which type of RNFE it is and therefore implies
that, pr(*RNFEp|DSj =1)>0 for j: mand pr(*RNFEp|D_{ =1)sofor j: f , where

s = hos,nhos and *RNFEp = RNFE,agRNFE,non - agRNFE .

HP4 There is no behavioral difference regarding non-farm participation
decisions among adult members of a PA death afflicted household from the time
following death shock to a later time period where the household could recover. This
hypothesis implicitly states that impacted households respond to the death shock but do
not recover from it regardless of what the gender of the deceased is. Time elapsed since a

PA death (TESPAD,) does not bring recovery because the household lacks resiliency. We

130

anticipate that when a death shock occurs, the household makes a base response but does
not change over time such that there will be no difference in decisions made among
households experiencing a recent male or female death (O-Syears) compared to those that

had an earlier male or female death (6-13 yrs ogo). Hence we can state that:

pr( RNFEpI TESPADtj ) = c1 , pr(agRNFEpI TESPADtj )= c2 and

pr(non —agRNFEp| TESPADtj )= c3, for j = m, f and cl,c2,c3 are constants. TESPAD,
represents either YSD! the discrete variable for the number of years since an afflicted
household suffered a prime age death or the binary variables; rpamj a household with

either a male or female recent PA deaths (O—5yrs) and epamj one with ether a male or

female earlier death (0-13yrs).

4.4.3 Level (Intensity) of Participation Hypotheses

HP5 Once individuals from afflicted households decide to participate in RNFE,
we anticipate that the level of hours worked will not be influenced more by the death
variables but by other controlling variables such as individual characteristics, farm
characteristics, household and market access factors. We generally hypothesize that death

variables do not determine the intensity of participation.

131

4.5 Empirical Methods and Data

4.5.1 Data and Econometric Model

Data

Data used in the regressions come from Malawi Agricultural Productivity and
Adult Mortality Survey of 2002. A total of 351 households sampled from representative
high maize production areas of Malawi’s Blantyre, Kasungu and Mzuzu Agricultural
Development Divisions (ADDs) are used in this analysis. Cross-sectional data and not
panel data were used for this study because the l989-90-baseline data set did not collect
information on intensity of non-farm labor time allocation. The study benefited from a
variable on cassava use prior to 2002, obtained from a 1997 survey conducted by

CIMMYT in the same study area.

Econometric Model

To examine how death stricken households participate in RNFE labor markets
and how their participation changes over an adjustment period where time has elapsed
from the initial death shock, we estimate participation and intensity level on aggregate

RNFE labor market as well as on agRNFE and non-agRNFE markets separately.

Participation Equations

First we run a regression for overall RNFE participation to examine how death

stricken households respond to the RNFE labor market. Since we argue that agRNFE and

132

non-agRNFE differ from each other in that the latter’s labor market has a higher barrier
to entry, we also run separate regression of Equation 28 below where
p = RNFEp,agRNFEp,non —agRNFEp is overall participation in RNFE and participation in

the agricultural and non-agricultural rural non-farm employment, respectively.

Pr] p = 1 ] = a+ flldeath var+ flzindividual var+ ,B3household var+ ,64 farm var+ ,Bsmarketaccess var+ E

(28)

The ,6 are regression coefficients associated with variables under each of the four vectors

of variables, a is a constant and e is the error term.

Intensity Equations

7 Intensity equations are also divided into three groups, overall RNFE, agricultural
RNFE, and non-agricultural RNF employment. The outcome variables are discrete
measures of hours worked per week by participating PA individuals where

h = RNFEhrs,agRNFEhrs,non —agRNFEhrs. The econometric equation is

h = ¢ + dldeath var+ 52individual var+ @household var+ 54 form var+ dsmarketaccess var+ ,u (29)

The 6 are regression coefficients associated with variables under each of the four groups

variables, to is a constant and p is the error term.

133

4.5.2 Empirical Model for participation in Aggregate RNFE, Agricultural and

Non-agricultural RNFE

The dependent variables for regression specification used toestimate individual
participation in overall RNFE, agricultural and non-agricultural RN FE markets in Malawi
are represented by RNFEp, agRNFEp and non-agRNFEp respectively. The variables
take on the value 1 when an individual participates and zero otherwise. It is assumed that
the likelihood of participation is affected by death variables, and by a set of four
controlling groups of variables; individual, household, farm, and location-market access
characteristics. All regressions in this essay use these explanatory variables shown in
Table 4.2, which correspond with the five categories identified from the general model.
We discuss these variables next beginning with death variables followed by other

controlling variables.
Death Variables

The death variables are binary variables that are defined by the existence of a PA
death in the household, the gender of the deceased (male or female) and his or her
position in the household as either a head or spouse (hos) or a non-head-or-spouse (nhos).
Four groups emerge from these; PA male hos death, PA female hos death, PA male nhos
death and PA female nhos death. To capture the timing aspects of the PA death shock
we split the death variables in two groups recent death 0—5 years (0/ 1) vs earlier death 6-
13 years (0/ l) interacted by the identity of the deceased and construct a discrete variable

YSD that captures the number of years since the death occurred. An attempt is

134

 

mks—«2 .95.. 50.: Saw mot—a ac gum—23 538.89. «30,—. 1:: 05.5: 93.59 05 3 «in: 333.3.» «0 333v. 0231809 N6 ozah

135

 

8.3:: 3.8. .3. 823% 221an .39 a... 2......

136

 

83.2.: .88. .9 erase 32.58 ea a... as:

137

NOON mE<m<2 U3.50m

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

wmvd womd 8 n 033550 :uv .8635 E2: o_o>oQ _E=:=o:m< Shale—m 353
556 >310 8 n ommauofio :nv c2230 Boga—goo _E2_:otw< :wcammm 338.
53 N. mac 0 Ex E Big: 8 8.835 Ego—E
when N36 380823 5 no; Sn 9 856mm 363
notatmtiato unwoouéeteood
mm m. : fio 8n oflauofio :nv 58.8 9 $88 .5232 2 320 3.253 E QEEBEo—z 228
032 :3“. «2“ 2:3 30,—. 29653
mm .3 cm .2 9200A: mgomma Boson—5; mo o=_m> mafia—E:
o0 «N mm.” can: 228:0: Co cosmos?» RES“ .8 £3» .8 LEE—dz 3813:

3. .mm Z.H.W _ 93... 205830; Co om< gag
wt. 3 Cd 8 n 33:28 :uv Eonomsoc venue; 295E o_mE&IEE
31.0 2.0 _ 0:8 Nocoegmon— 2:: 0595 on
who; mm QM o>onm 95 mac 2 2323?: :35 me 59:52 :35
83.22.6335» 32th::

Svd cmNd Emit/850 n o ”8.5 m 238:0; n 3 3:0 m 29.850: .2 =32sz lasteda
v2.0 owmd 823550 n 0 ”can: 20:839. u C can: 30:83: w_ :32an 9321::
mhmd god 33E u o ”Bessy n C oEEom £ 391% “onEoE 20:38: :32an Beacon

55359

c.5235 :32 553.689 0322:,

 

 

83%: 3.2.. no “Page“ 33:83 .39 a... 05.;

138

also made to further subdivide death groups into four categories 0-3yrs; 4—7yrs; 7-9yrs;
10-13yrs to pinpoint short-term, medium, long and very long-term response.

A small but growing number of studies have assessed the effect of adult death on
non-farm labor allocation decisions. Beegle (2003) used panel data to study the effect of
adult death on intra-household labor substitution in Tanzania, while Anglewicz (2005)
extended Beegle’s (2003) approach to incorporate morbidity effects as well as the
mortality effects. Use of panel data enabled them to track households and control for
individual and household fixed effects. Anglewicz (2005) found that HIV/AIDS-related
illness and death induced diversification of income sources, with women reallocating
time from work-intensive farming to cash-generating employment (ganyu). Beegle
(2003) on the other hand found participation in wage employment was not compromised
by past PA adult death, but reduced in the short-run period before death. Non—farm self-
employment participation had no significant changes associated with male or female
deaths.

These studies show that adult death impacts could go either way depending on
gender of the deceased, tinting before or after death and location of the study. Adults in a
household that suffered a prime age death could decide on not participating in the non-
farm sector to consolidate farming labor and to maintain agricultural output given a loss
of labor. In this case the time elapsed between the death and our survey timing in 2002
may be crucial with individuals eventually participating after dealing with post-death
adjustment. On the contrary, a household suffering a PA death may opt to participate in
the off-farm sector following a financial drain from illness and funeral expenses. Non-

farm cash employment has quicker payoffs than farming, even of annual crops. This

139

“financial push” effect could likely induce individuals to participate initially and
eventually reduce level of participation with increased time elapsed since death, as the
Beegle (2003) study suggests. The death effect is also dependent on the gender and
household position of the deceased. If the deceased was mOre of a financial contributor,
for instance, the case of someone who contributed to cash cropping or remitted finances,
we are likely to see the “financial push” effect. If the deceased was more of a household

labor contributor, we are more likely to observe a “labor pull” effect.

Individual characteristics

Individual characteristics included in the regression are age, education, the
household position and gender. We expect younger adults to be more likely to participate
than older adults as they have more mobility, agility and opportunity. Educated adults
are more likely to be involved in non—agricultural non-farm activities, as they are more in
tune with options outside of agriculture. We expect women to be less likely to be
involved in non-agRNFE because of capital constraints and the challenge of balancing
household chores and agricultural production with non-farm activities. Zwick and Smith
(2001) found reproductive labor demands, weak property rights and strong traditions to
be barriers for women’s participation in all income-generating project in Rakai District,
Uganda. Spouses of household heads carry other household responsibilities and are
hence less likely to be involved in RNFE activities while other members (non-
head/spouse) have fewer domestic responsibilities and are therefore more likely to be

engaged in non-farm employment.

140

Household Characteristics

The total number of adults in the household is normally associated with household
capacity to diversify its income generating activities so that the larger the number of
adults, the greater the probability of individual adult participation in non-farm
employment (Matshe and Young, 2004). We expect larger households to have more
individuals participating in RNFE.

Dependency ratio, computed as the proportion of non-prime age family members
(anyone less than 15years old or 60 years and older) to prime age members (15-59 years
old) is expected to be a key determinant to individual participation. The higher this ratio,
the more likely individuals will be involved in non-farm employment to supplement
household income. We expect individuals from female-headed households to be more
likely involved in non-farm work because they may not benefit from other farm
production enhancement opportunities like input credit facilities that are mainly
acceSsible only to men. Education level and age of the head of household are expected to
increase individual participation. The value of housing assets is expected to have an
inverse relationship with decision to participate. Those households with good housing
may be wealthy enough to not bother to engage in non-farm sector compared to those
from poor households with low value housing. Bryceson (2006) found poor households
participating in ganyu to be an entrenched feature of many local rural economies in

Malawi.

141

Farm Characteristics

We anticipate individuals from households that own more land to be less likely to
participate in non-farm employment because they generate enough earnings from farming
to not bother with supplementary income. Cassava production has been mentioned as a
possible coping mechanism for PA death afflicted households trying to cope with reduced
labor in the literature (Rugalema, 1983, Barnett and Blaikie, 1992).

Individuals who participate in RN FE could use credit to have capital outlay for
non-agRNFE. Credit in rural Malawi is accessed through farmers’ clubs that help secure
it through group lending. Club membership could therefore explain participation in
RNFE. Club membership however has ambiguity to it. It could imply access to credit in
which case individuals have access to inputs for profitable agricultural production that
generate enough revenue and hence no need for extra earnings. On the contrary,
households with credit may be willing to participate in non-farm labor markets in order to
generate more income for debt servicing. Matshe and Young (2004) did not find credit a

significant determinant of participation.

Location and Market-access Characteristics

Location and market access characteristics capture market access factors that
influence relative profitability of involvement in non-farm work and the terms of trade
between farm and non-farm sectors. Such factors as crop prices, wages in different
sectors and transportation cost belong to this group. Unfortunately, these data were not

collected in the survey. Distance to market and transport are used a proxies. Escobal

142

(2001) found distance to market highly significant (1%) and negative as a determinant of
self-employment in non-agricultural activities but less significant (10%) and negative as a
determinant of wage employment skilled non-agricultural activities. We expected a
significant and negative response to this variable also because the greater the distance the
less the net income gains because of transportation costs. The large transaction cost will
deter many from engaging in distant RNFE activities and hence non-agRNFE.

The Southern region of Malawi is home to the city of Blantyre where most
manufacturing and commerce occur. We expect individuals located in Blantyre ADD to
be more likely involved in non-farm activity, followed by Kasungu in Central Malawi
and less likely if located in Mzuzu in the North. There could be more agricultural RNFE
employment in the Northern region due to larger farm size. Regional dummy variables
also capture social characteristics such as marital systems. Mzuzu is predominantly
patrilineal3 while Kasungu and Blantyre ADD are matrilineal“. Men may be more
involved in non-farm work in patrilineal societies and women more in matrilineal
according to power relations associated with headship of the household. Kerr (2005)
found men more involved in casual farm labor ganyu, in the mainly patrilineal Northern

region because of gender power bias.

 

3 Descent and inheritance follow the paternal (father’s) lineage
4 Descent and inheritance follow the maternal (mother’s) lineage

143

4.5.3 Empirical Model for Level of Intensity of Participation in Agriculture and

Non-agricultural RNFE

As per our hypothesis, once a person has chosen to’participate in RNFE, the
intensity level or the number of hours worked per week could potentially depend on all
the same factors or a subset of those factors. Only by conducting an empirical analysis
will we be able to ascertain this. The dependent variables for regression specification
used to estimate individual intensity of participation in agricultural and non-agricultural
RNFE markets in Malawi are represented by agRNFEhrs and non-agRNFEhrs

respectively.

4.5.4 Econometric Concerns

First, there is concern over choosing the underlying decision making process,
whether joint or separate as mentioned in the modeling section of this essay. We will run
the Likelihood Ratio Test, to choose between the Tobit model and the double hurdle
model.

Second, the regression model uses data collected at different scales (individual,
household and plot level) that could potentially result in heteroskedasticity problems in
the error term (Wooldridge, 2002). We will estimate the models using heteroskedasticity
robust (White- corrected) standard errors to forestall this problem.

Third, there is a potential simultaneity bias problem in using the variable
“household growing cassava in 2002”, the year the survey was conducted. Individuals

deciding to participate in RNFE could also be making this decision jointly with growing

144

cassava in that year because it saves labor and therefore releases individuals’ time to
engage more in off-farm employment. We will conduct a bivariate probit regression to
test the hypothesis that there is no simultaneity between growing cassava in 2002 and

participating in RNFE in 2002 (ie p: 0). If the null is rejected we will use a variable that

captures prior cassava production to avoid potential simultaneity bias.

Fourth and last, there are potential multicollinearity problems given we are using
dummy variables. Related to this issue is perfect correlation between dependent
variables and independent variables. This is likely to happen with timing of death
variables. Mortality figures in a small sample are generally low. Further, disaggregating
death variables into two or four periods could reduce degrees of freedom in each period
leading to perfect predictability in which death variables could be dropped from
regression. We will run a correlation matrix and variance inflation factor to assess the

severity of correlation and multicollinearity if any.

4.6 Regression Results

4.6.1 Results on Univariate Analysis of Types of Non-farm Activities

There were several types of non-farm activities in which individuals from rural
Malawian households engaged. Table 4.3 shows the ranking of the most popular
activities split between agricultural and non-agricultural. Casual farm labor (Ganyu) is
the major agricultural non-farm employment activity, followed by Dimba gardening on
moist flood recessed land and working in estates providing agricultural labor. Beer

brewing and buying and selling (Gerri) rank first and second among non-agricultural

145

Rural Non-Farm Employment. Contract work, fishing and carpentry are ranked third,
forth and fifth respectively. Various other activities were reported from the survey,
including marketing of fruits, sugar cane and firewood, tailoring, tinsmith, pottery,

basketry, mat making, tobacco trading, cattle trading and bee keeping as shown in Table

 

 

 

 

4.3 below.
Table 4.3 Adult individual participation in types of non-farm employment activities available in
rural Malawi, 2002
Type of Activity No of Adults Participants involved Percent
AgRNFE
Casual labor (Ganyu) 79 73.8
Dimba Gardens 18 16.8
Estate Labor 7 6.5
Draft power hire 3 2.8
Total 107 100.0
Non-agRNFE
Beer Brewing 38 23.0
Buying and selling (geni) 25 15.2
Contract work 20 12. 1
Fishing 10 6.1
Carpentry 9 5.5
Selling Fruits (& sugar cane) 7 4.2
Selling Firewood 7 4.2
Basketry 5 3.0
Pottery 4 2.4
Wood cutting 4 2.4
Tobacco Trading 4 2.4
Tailor 4 2.4
Builder 4 2.4
Uniform making 3 1.8
Tinsmith 3 1.8
Brick molder 3 1.8
Migrant worker 3 1.8
Bee keeping 2 1.2
Mat making 2 1.2
Forestry Department worker 2 1.2
Small business 2 1.2
Cattle Trading 1 0.6
Grass mowing l 0.6
Herbalist l 0.6
Handicraft 'l 0.6
Total 165 100.0

 

Source: MAPAMS 2002

146

Participation by Gender

Participation by gender is shown in Table 4.4. Casual labor and dimba gardening
are roughly equally participated in by both genders among agricultural rural non-farm
employment activities. Estate wage labor and draft power hire predominantly man’s

domain. Beer brewing is largely a woman’s non-agRNFE while buying and selling,

Table 4.4 Individual adult participation in main non-farm employment activities by gender in

 

 

 

 

rural Malawi, 2002

Type of Activity Male (%) Female (%) Number of
participants

AgRNFE
Casual labor (Ganyu) 53.2 46.8 79
Dimba Gardens 50.0 50.0 18
Estate Wage Labor 85.7 14.3 7
Draft Power Hire 66.7 33.3 3
Non-agRNFE
Beer Brewing 7.9 92.1 38
Buying and selling (geni) 72.0 28.0 25
Contract work 75.0 25.0 20
Fishing 80.0 0.0 10
Carpentry 100.0 0.0 9
Selling Fruits (& sugar cane) 57.1 42.9 7
Selling Firewood 42.9 57.1 7
Basketry 60.0 40.0 5
Pottery 50.0 50.0 4
Wood cutting 100.0 0.0 4
Tobacco Trading 50.0 50.0 4
Tailor 5 0.0 50.0 4
Builder 100.0 0.0 4
Uniform making 66.7 33.3 3
Tinsmith 100.0 0.0 3
Brick molder 66.7 33.3 3
Migrant worker 100.0 0.0 3
Bee keeping 100.0 0.0 2
Mat making 100.0 0.0 2
Forestry Department worker 100.0 0.0 2
Small business 100.0 0.0 2
Cattle Trading 100.0 0.0 1
Grass cutting 0.0 100.0 1
Herbalist 100.0 0.0 l
Handicraft 1 00.0 0.0 1
Total 272

 

Source: MAPAMS 2002

147

contract work; fishing and carpentry tend to be largely dominated by men. Overall the

skewed participation in Table 4.4 reveals clear gender roles.

Non-farm employment participation by region

The availability of non-farm employment opportunities in rural Malawi is shown
in Table 4.5. Kasungu Agricultural Development Division and Mzuzu ADD have higher
percentages of ganyu casual labor (above 40 percent) participants compared to Blantyre
ADD. Kasungu Agricultural Development Division (ADD), located in the Central
Province of Malawi, appears to lead in estate wage labor while Mzuzu ADD leads in
Dimba Gardens. Buying and selling, contract work, selling of fruits and firewood in
Kasungu constitute 50% or more of RNFE participants in the survey. This is possibly
because Kasungu is nearer the capital city of Lilongwe, and that ADD is one of Malawi’s
major farming areas with good road networks. Carpentry, Tobacco trading, Tailoring and
builder represent the four non-agRNFE activities in which Blantyre ADD has the highest
proportion of participants among all three ADDs. The city of Blantyre, known as the
commercial capital of Malawi, is located in the ADD of the same name. The high
population density of the south makes marketing ventures a viable option for most rural
individuals. Mzuzu, in the Northern Region of Malawi has highest proportion of beer
brewing, basketry and pottery activities. Mzuzu ADD, like most of the Northern region,
has low population density and inadequate infrastructure that restricts livelihood options

to localized activities such as beer brewing and casual labor.

148

Table 4.5 Individual adult participation in main non-farm employment activities by region in

 

 

 

 

rural Malawi, 2002
Type of Activity % of % of % of Number
Participants in Participants in Participants of participants
Blantyre ADD Kasungu ADD Mzuzu
ADD
AgRNFE
Casual labor (Ganyu) 15.2 40.5 44.3 79
Dimba Gardens 0.0 33.3 66.7 18
Estate Wage Labor 14.3 57.1 28.6 7
Draft Power Hire 0.0 100.0 0.0 3
Non-agRNFE
Beer brewing 18.4 23.7 57.9 38
Buying & Selling (Geni) 41.7 50.0 8.3 24
Contract Work 10.0 65.0 25.0 20
Fishing 50.0 30.0 20.0 10
Carpentry 66.7 22.2 1 1.1 9
Selling Fruits (& sugar cane) 28.6 71.4 0.0 7
Selling Firewood 14.3 85.7 0.0 7
Basketry 40.0 0.0 60.0 5
Pottery 0.0 0.0 100.0 4
Wood cutting 50.0 0.0 50.0 4
Tobacco Trading 75.0 25.0 0.0 4
Tailor 100.0 0.0 0.0 4
Builder 75.0 0.0 25.0 4
Uniform making 0.0 33.3 66.7 3
Tinsmith 0.0 33.3 66.7 3
Brick molder 33.3 33.3 33.3 3
Migrant worker 33.3 33.3 33.3 3
Bee keeping 0.0 0.0 0.0 2
Mat making 0.0 0.0 100.0 2
Forestry Department worker 50.0 0.0 50.0 2
Small business 50.0 0.0 50.0 2
Cattle Trading 0.0 100.0 0.0 1
Grass cutting 0.0 100.0 0.0 1
Herbalist 0.0 100.0 0.0 l
Handicraft 0.0 100.0 0.0 1
Total 272

 

Source: MAPAMS 2002

149

Non-farm Income by region

Mean per capita non-agricultural rural non-farm income is over three times the
average per capita income in Malawi Kwacha (MK) earned from agricultural income,
that is, MK 3,164 compared to MK 9,616. Comparing across gender lines income from
non_agRN FE is higher than that from ag_RNFE. Women earned MK 3,705 from
non_agRNF income compared to MK 3,407 from ag_RNFE. Per capita income from
ag_RNFE among the men was lower than women’s per capita average income at MK
2,963. However, the income from non_agRN FE was more than three times the amount
of MK 13,679 earned by women. Across ADDS, non_agricultural rural incomes were
higher than incomes from agriculture. Male per capita mean incomes were generally
higher than female per capita incomes with the exception of Kasungu ADD where

agricultural incomes are higher non-agricultural incomes as shown in Figure 4.1.

150

 

 

oz_-mu_zmm<-coz I
oz_-mu_zmu<l

  

CIGVZW 101 “991/11

   

non< use ..oucom >2 2:005
333th mug—mafia: new mmzmmu .35..“ :32 ...e 959".

919U19:l 'ClClVZlN
919W 'ClCIVZW
amil-19:1 ' CICIV>|

00W 10; ueew
GOV/18 10! UBGW
aI91119:l ' CIGV'IG

919W ‘ GOV/)1
919W ' OClV'lQ

 

0:5 0:32.. 233th muzm .m:::< cams.

ooom

,ar oooow

ooomw

oooom

000mm

6m) Buoemx werew

 

 

151

Group Comparison: Test of Means between Participants and Non-participants in RNF E

Table 4.6 gives summary statistics and t-tests of equality of means for variables
used in estimating participation and intensity of participatiOn models. We test the null
hypothesis that the means of participant and non-participant adults are equal against the
alternative that the means are not equal. The results in Table 4.6 give the overall
participation in aggregate RNFE.

There are a number of significant differences between RNFE participants and
non-participants. The mean death variables for participants are significantly higher than
those of non-participant group for both recent and earlier death afflicted households.
When death timing variables are broken down into four categories only the 10- 13 years
after PA death variables shows significant differences. Adult individual who participate
in RNFE are more likely to come from households that have had a death of male or
female spouse 10-13 years previously. Adults who participate in RNFE are older and
more likely to be household heads than non-participants. Women are less likely than
men to participate in RNFE. At the household level, participants come from households
with higher dependency ratios, older and more educated headed households that grow
cassava and had a male death of a household head or spouse. The participants also come
from households with significantly fewer adults and poorer housing than non-participant
households. Participants came out of significantly fewer households that grew cassava in
1997 but significantly more households that grow cassava during the survey 2002

compared to non-participants.

152

Table 4.6 Comparative statistics of participants and non-participant individual’s data used in

empirical estimations of rural non-farm employment participation (RNFEp) in Malawi,

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2002
Participants Non-Participants Test of Means
Variable (n—270) (n—849) (non p art)- ( p art)a
Mean 1 Std. Dev Mean 1 Std Dev t-stat I p-val
Individual Characteristics
Age (yrs) 43.48 14.99 33.79 18.72 -7.74 0.00
Age? (“52) 2114 1363 1493 1639 -5154 0.00
Education (yrs) 5.12 4.03 5.74 3.72 2.35 0.02
Female (0,1) 0.42 0.49 0.53 0.50 3.11 0.00
Household head (0,1) 0.60 0.49 0.19 0.39 -l3.84 0.00
Household spouse (0,1) 0.21 0.41 0.24 0.46 1.01 0.31
Household Characteristics
Adults (Number) 3.30 1.46 4.22 2.06 6.82 0.00
Dependency ratio 1.12 0.86 1.01 0.84 -l .79 0.07
Female Headed (0,1) 0.16 0.37 0.06 0.24 -5.17 0.00
Head’s Age (yrs) 29.3 25.9 10.76 23.27 -11.10 0.00
Head’s Education (yrs) 2.98 4.07 0.75 2.08 -1 1.82 0.00
Housing Assets (‘000MK) 8.83 19.15 14.57 31.54 2.82 0.01
Land holding (Ha) 1.37 0.99 1.47 1.20 1.26 0.21
Grow Cassava (0,1) 0.12 0.32 0.08 0.28 -1.73 0.08
Grow Cassava 1997 (0,1) 0.07 0.27 0.11 0.31 1.87 0.06
Farmer Club member(0, 1) 0.09 0.29 0.13 0.33 1.49 0.14
gcation-access variables
flantyre ADD (0,1) 0.24 0.43 0.32 0.47 2.52 0.01
igsungu ADD (0,1) 0.43 0.50 0.32 0.47 -3. 14 0.00
Mzuzu ADD (0,1) 0.34 0.47 0.36 0.48 0.71 0.48
(fistance to Bus Stop km 5.46 6.44 5.95 6.75 1.04 0.30
ktance to market (km) 5.94 8.13 6.06 7.70 0.21 0.83
_!:fz_°me Age Death Variables for Households (0,1)
& death 0.22 0.42 0.19 0.39 -l .29 0.20
& male death 0.15 0.36 0.1 l 0.31 -l.88 0.06
_& female death 0.1 1 0.31 0.10 0.30 -0.06 0.95
£male hos death 0.09 0.29 0.06 0.23 -2.17 0.03
LPA‘female hos death 0.05 0.22 0.05 0.23 0.15 0.88
tprmale nhos death 0.06 0.23 0.06 0.24 0.28 0.78
EL female nhos death 0.06 0.23 0.05 0.22 -0.24 0.82
kars elapsed since a Prime age death (years)
.3ar Since Death (YSD) 1 0.66 1 2.00 T017 [ 1.04 1 -523 1 0.00

 

 

SOurce: MAPAMS 2002

hos = female head or spouse, nhos = non head or spouse

Difference in Means = (mean non-participants - mean participants)

153

Table 4.6 (cntd). Comparative statistics of participants and non-participant individual’s data used in
empirical estimations of rural non-farm employment participation (RNFEp) in Malawi,

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2002
Participants Non-Participants Test of Means

Variable (n-270) (n—849) (non p art)— ( p art)a

Mean 1 Std. Dev Mean 1 Std Dev t-stat [ p—val
Prime Age Death timing variables for Households (0, 1)
PA death past 0—3yrs 0.11 0.31 0.10 0.30 -0.34 0.74
PA death past 4-6yrs 0.07 0.26 0.05 0.22 -1.15 0.25
PA death past 7-9yrs 0.05 0.21 0.05 0.22 0.24 0.81
PA death past 10-l3yrs 0.03 0.17 0.01 0.1 l -2.03 0.04
PA mle death past 0-3yrs 0.06 0.24 0.04 0.19 -l .42 0.16
PA fml death past 0-3yrs 0.05 0.22 0.07 0.25 1.06 0.27
PA mle death past 4-6yrs 0.04 0.21 0.03 0.18 -0.78 0.44
PA fml death past 4-6yrs 0.03 0.17 0.02 0.14 -O.80 0.43
PA mle death past 7-9yrs 0.03 0.18 0.04 0.19 0.42 0.68
PA fml death past 7-9yrs 0.02 0.12 0.01 0.1 1 -0.23 0.82
PA mle death 10-13yrs 0.02 0.12 0.01 0.08 -1.43 0.15
PA fml death 10-13yrs 0.02 0.12 0.01 0.08 -l.43 0.15
PA mhos death 0-3yrs 0.05 0.22 0.03 0.17 -1.36 0.17
PA fhos death 0-3Jrs 0.05 0.22 0.03 0.17 -1.36 0.17
PA mhos death 4-6yrs 0.05 0.22 0 .04 0.20 -0.57 0.57
PA fhos death 4-6yrs 0.05 0.22 0.04 0.20 -0.57 0.57
PA mhos death 7-9yrs 0.02 0.15 0.03 0.16 0.24 0.81
PA fhos death 7-9yrs 0.02 0.15 0.03 0.16 0.24 0.81
PA mhos death 10-l3yrs 0.02 0.15 0.01 0.09 -1.87 0.06
PA fhos death 10-l3yrs 0.02 0.15 0.01 0.09 -l.87 0.06
PA mnhos death 0-3yrs 0.04 0.19 0.04 0.19 0.05 0.96
PA fnhos death 0-3yrs 0.04 0.19 0.04 0.21 0.55 0.58
PA m-nhos death 4—6yrs 0.01 0.09 0.02 0.12 0.99 0.33
PA f-nhos death 4—6yrs 0.02 0.12 0.01 0.08 -1.42 0.15
PA m-nhos death 7-9yrs 0.02 0.15 0.03 0.17 0.54 0.59
PA f-nhos death 7-9yrs 0.02 0.12 0.01 0.09 -0.95 0.34
PA m-nhos death 10-l3yrs 0.00 0.06 0.00 0.05 -0.37 0.71
PA f-nhos death 10-13yrs 0.01 0.09 0.00 0.05 -1.21 0.23

 

 

 

 

 

 

 

Source: MAPAMS 2002

fhos 2 female head or spouse, mhos = male head or spouse
f-nhos = female non-head or spouse household member,
m-nhos = non-head or male other household member

a . . . . . .
Drfference 1n Means = (mean non-partrcrpants - mean parttcrpants)

154

Table 4.6 (cntd) Comparative statistics of participants and non-participant individual’s data used in
empirical estimations of rural non farm employment participation (RNFEp) in Malawi,

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2002
Participants Non-Participants Test of Means
' = = 3

Variable (n 270) (n 849) (non pa rt)- ( p art)

Mean rStd. Dev Mean I Std Dev t-stat I p—val
Prime Age Death timing (Recent[ 0-5] vs Earlier deathsI6-l3]) variables for Households (0,1 1
Timing of death
PA recent death 0-5yrs 0.10 0.31 0.02 0.14 -6.36 0.00
PA earlier death 6-l3yrs 0.05 0.22 0.02 0.12 -3.43 0.00
Gender by timing of death
PA mle rec. death 0-5yrs 0.06 0.23 0.01 0.10 -4.70 0.00
PA fml ear. death 0-5yrs 0.05 0.22 0.01 0.10 —4.11 0.00
PA mle rec. death 6—13yrs 0.03 0.16 0.01 0.10 -1.85 0.07
PA fml ear. death 6—13yrs 0.03 0.16 0.01 0.07 -3.09 0.02
Head or spouse deaths
PA mle hos death 0-5yrs 0.04 0.19 0.00 0.06 -4.51 0.00
PA fml hos death 0-5yrs 0.02 0.12 0.01 0.08 -1.18 0.24
PA mle hos death 6-13yrs 0.02 0.12 0.01 0.08 -1.43 0.15
PA fml hos death 6-13yrs 0.02 0.16 0.01 0.07 -2.67 0.01
Non-head or spouse deaths
PA mle nhos death 0-5yrs 0.02 0.12 0.01 0.08 -l.92 0.06
PA fml nhos death 0-5yrs 0.03 0.01 0.00 0.05 -4.53 0.00
PA mle nhos death 6-13yrs 0.01 0.11 0.01 0.07 -1.16 0.25
PA fml nhos death 6-13yrs .001 0.03 0.00 0.00 -l.78 0.08

 

 

 

 

 

 

 

 

 

 

Source: MAPAMS 2002

Notes

rec. = recent. ear. = earlier.

fml hos = female head or spouse, mle hos = male head or spouse
fml nhos = female non-head or spouse, mle-nhos = non-head or spouse

a . . . . . .
Difference 1n Means = (mean non-partrcrpants — mean partrcrpants)

155

In terms of location and market access, there are significantly more RNFE
participants in Kasungu ADD than non-participants, while in Mzuzu ADD the opposite is
true. For Mzuzu ADD there is no significant difference between participants and non-

participants.

Group Comparison: Test of Means between Participants in agRNFE and Non-agRNF E

We now turn to look at the subsets of RNFE: agricultural versus non-agricultural.
Table 4.7 suggests a few minor differences between participants in agricultural and non-
agricultural RNFE regarding death variables. There are more participants from
households afflicted with recent or earlier death of men in the non-agRNFE than in the
agRNFE. The pattern is noted for households with recent death of a woman head or
spouse.

The test of means between participants in agricultural and non-agricultural RNFE
shows no significant difference among individual characteristics except for age.
Participants in non-agRNFE are significantly older. At the household level, results show
that participants in agricultural RNFE come more from female-headed households and
that the age and education levels of these heads of households are significantly less than
those from households in which individuals participate in non-agRNFE. For location and
market access, there is significant difference between participants in the two RNFE

markets. Mzuzu ADD has significantly more agRNFE participants, while Blantyre has

156

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.7 Comparative statistics of participants in AgRNFE and Non-agRNFE for individual’s
data used in empirical estimations of rural non farm employment participation
(RNFEp) in Malawi, 2002
AgRNFE Non-agRNFE Test of Means
Variable Participants Participants
(""0” (“"6” tag-part)—(nag-part)a
Mean IStd. Dev Mean I Std Dev t-stat I p-val
Individual Characteristics
Age (yrs) 40.30 15.20 45.50 14.50 -2.69 0.00
Agez (”52) 1867 1303 2278 1380 —2.48 0.01
Education (yrs) 4.90 3.90 5.24 4.10 -0.63 0.53
Female (0,1 ) 0.45 0.50 0.41 0.49 0.69 0.49
Household head (0,1) 0.54 0.50 0.63 0.48 -l.44 0.15
Household spouse (0,1) 0.76 0.43 0.84 0.37 -1.57 0.12
Household Characteristics
Adults (Number) 3.24 l .42 3.35 1.45 -0.60 0.55
Dependency ratio 1.15 0.92 1.09 0.82 0.53 0.60
Female Headed (0,1) 0.2] 0.41 0.13 0.33 1.66 0.10
Head’s Age Qrs) 25.58 25.16 31.72 26.04 -1.94 0.05
Head’s Education (yrs) 2.35 3.30 3.38 4.47 -2.20 0.03
Housing Assets (‘000MK) 10.90 21.10 7.73 18.01 1.29 0.20
Land holding (Ha) 1.38 1.11 1.30 0.89 0.64 0.52
Grow Cassava (0,1) 0.15 0.36 0.10 0.30 1.26 0.21
Grow Cassava 1997 (0,1) 0.10 0.31 0.06 0.23 1.40 0.16
Farmer Club member (0,1) 0.07 0.26 0.10 0.31 -0.81 0.42
Location-access variables
Blantyre ADD (0,1) 0.12 0.33 0.31 0.47 -3.90 0.00
Kasungu ADD (0,1) 0.39 0.49 0.44 0.50 -0.81 0.42
Mzuzu ADD (0,1) 0.49 0.50 0.25 0.43 4.02 0.00
Distance to Bus Stop km 5.31 6.80 5.57 6.19 —0.32 0.75
Distance to market (km) 6.54 8.85 5.62 7.60 0.88 0.38
Prime Age Death Variables for Households (0,1)
PA death 0.61 0.37 0.26 0.44 -2.06 0.04
PA mle death 0.12 0.33 0.16 0.37 -0.98 0.33
PA fml death 0.08 0.28 0.12 0.33 - 1 .00 0.32
PA mle hos death 0.1 1 0.32 0.08 0.27 0.90 0.37
PA fml hos death 0.03 0.17 0.07 0.25 -l .53 0.13
PA mle nhos death 0.01 0.10 0.08 0.28 -3.19 0.00
PA fml nhos death 0.06 0.23 0.05 0.23 0.05 0.96
Years elajged since a Prime age death (years)
Year Since Death (YSD) I 0.54 I 1.77 I 0.72 2.14 I -075 I 0.45

 

Source: MAPAMS 2002
rec. recent.

fml hos = female head or spouse,

fml nhos 2 female non-head or Spouse,

ear. =earlier.

mle hos = male head or spouse
mle-nhos = non-head or Spouse

a
Difference in Means = (mean agRNFE-participants — mean non-agRNFE participants)

 

Table 4.7 (cntd) Comparative statistics of participants in AgRNF E and Non-agRNFE for
individual’s data used in empirical estimations of rural non farm employment
participation (RNFEp) in Malawi, 2002

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

AgRNFE Non-agRNFE Test of Means
Variable Participants Participants

(n—105) (n—l65) (ag-part)-(nag-part)a

Mean I Std. Dev Mean I Std Dev t-stat I p-val

Prime Age Death timing variables for Households (0,1)
PA death past 0-3yrs 0.08 0.28 0.12 0.33 -1.00 0.32
PA death past 4-6yrs 0.07 0.25 0.07 0.26 -0.23 0.82
PA death past 7-9yrs 0.03 0.17 0.06 0.24 -1.33 0.17
PA death past 10-13yrs 0.03 0.17 0.03 0.17 -0.1 1 0.91
PA mle death past 0—3yrs 0.04 0.19 0.07 0.26 -1.29 0.20
PA fml death past 0-3yrs 0.05 0.21 0.05 0.25 -0.07 0.95
PA mle death past 4-6yrs 0.05 0.21 0.04 0.20 0.17 0.87
PA fml death past 4-6yrs 0.02 0.14 0.04 0.19 -0.90 0.37
PA mle death past 7-9yrs 0.02 0.14 0.04 0.20 -l .16 0.25
PA fml death past 7-9yrS 0.01 0.10 0.02 0.13 -0.63 0.53
PA mle death 10-l3yrs 0.02 0.14 0.01 0.11 0.42 0.68
PA fml death 10-13yrs 0.01 0.10 0.02 0.13 -0.63 0.53
PA mhos death 0-3yrS 0.05 0.21 0.05 0.21 -0.07 0.95
PA fhos death 0-3yrs 0.05 0.21 0.05 0.24 -0.07 0.95
PA mhos death 4-6yrs 0.06 0.23 0.04 0.20 0.50 0.62
PA fhos death 4—6yrS 0.06 0.23 0.04 0.20 0.50 0.62
PA mhos death 7-9yrS 0.02 0.14 0.02 0.15 -0.32 0.75
PA fhos death 7-9yrs 0.02 0.14 0.02 0.15 -0.32 0.76
PA mhos death 10-13yrs 0.02 0.14 0.02 0.15 -0.32 0.76
PA fhos death 10-13yrs 0.02 0.14 0.02 0.15 -0.32 0.76
PA m—nhos death 0-3yrs 0.00 0.00 0.06 0.24 -3.25 0.00
PA f-nhos death 0-3yrs 0.05 0.21 0.03 0.17 0.67 0.50
PA m—nhos death 4-6Lrs 0.00 0.00 0.01 0.1 1 -l.42 0.16
PA f-nhos death 4-6yrs 0.01 0.10 0.02 0.13 -0.63 0.53
PA m—nhos death 7-9yrs 0.00 0.00 0.04 0.19 -2.49 0.01
PA f-nhos death 7-9yrs 0.02 0.14 0.01 0.11 0.42 0.68
PA m-nhos death 10-13yrs 0.01 0.10 0.00 0.00 1.00 0.32
PA f-nhos death 10-13yrs 0.00 0.00 0.01 0.1 l -l .42 0.16
Source: MAPAMS 2002
rec. recent. ear. =earlier.
fml hos = female head or spouse, mle hos = male head or Spouse
fml nhos = female non-head or Spouse, mle-nhos = non-head or Spouse

a
Difference in Means = (mean agRNFE-participants — mean non-agRNFE participants)

158

 

Table 4.7 (Ctnd) Comparative statistics of participants in AgRNFE and Non-agRNF E for
individual’s data used in empirical estimations of rural non farm employment
participation (RNFEp) in Malawi, 2002

 

 

AgRNFE Non-agRNFE Test of Means
Variable Participants Participants
(n=105) (n=207)

a
tag-pam-(nag-pan)

 

 

 

 

Mean I Std. Dev Mean I Std Dev t-stat I p-val

 

Prime Age Death timing (Recentl 0-5] vs Earlier deaths] 6-1 3 I ) variables for Households (0,1)

 

Timing of death

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PA recent death 0-5yrs 0.08 0.28 0.12 0.32 -0.85 0.40
PA earlier death 6-13yrs 0.04 0.10 0.06 0.24 -0.87 0.38
Gender by timing of death

PA mle rec. death Gigs 0.05 0.21 0.06 0.24 -0.50 0.62
PA fml ear. death O-SJrS 0.04 0.19 0.05 0.23 -0.67 0.50
PA mle rec. death 6-l3yrs 0.02 0.14 0.03 0.17 -0.62 0.54
PA fml ear. death 6-13yrS 0.02 0.14 0.03 0.17 -0.62 0.54
Head or spouse deaths

PA mle hos death 0-5yrs 0.05 0.21 0.03 0.17 0.67 0.50
PA fml hos death 0-5yrs 0.00 0.00 0.02 0.15 -2.02 0.05
PA mle hos death 6-13yrs 0.02 0.14 0.01 0.11 -2.26 0.03
PA fml hos death 6- ] 3yrs 0.02 0.14 0.02 0.15 0.31 0.76
Non-head or spouse deaths

PA mle nhos death 0-5yrs 0.00 0.00 0.03 0.17 0.42 0.68
PA fml nhos death 0-5yrs 0.04 0.19 0.03 0.17 -0.31 0.76
PA mle nhos death 6-l3yrs 0.00 0.00 0.02 0.13 -l .74 0.08
PA fml nhos death 6-13yrs 0.00 0.00 0.01 0.08 -l.00 0.32

 

 

 

 

 

 

 

 

 

Source: MAPAMS 2002

rec. = recent. ear. = earlier.
fml hos 2 female head or Spouse, mle hos = male head or spouse
fml nhos = female non-head or spouse, mle-nhos = non-head or spouse

3
Difference in Means = (mean agRNFE-participants — mean non-agRNFE participants)

159

 

Significantly more non-agRNFE participants. This is probably an indication of
availability of agricultural types of jobs given the high per capita land ratio in Mzuzu

ADD while Blantyre ADD, home of Blantyre City, has more non—agricultural jobs.

4.6.2 Results on Tobit vs Double Hurdle Model

Tobit versus Double Hurdle Model

Using the test suggested by Greene (2002) we computed the LR test statistic,

lambda = -2(LLprobit + LLtruncated — LLTobit) = 273 and compared it to Chi-square

value 1.75 22 = 33.92. (see Appendix C, Tables C1 and C2). We reject the null hypothesis

HP. that claims that the restricted model (Tobit) is better than the double hurdle model.
We therefore conclude that RNFE participation and intensity decisions are separate rather
than joint. Because of this conclusion, discussion of regression results will mainly focus
on the double hurdle model approach — using probit estimation for participation and

truncated regression for hours worked per week.

Other Econometric Concerns

Heteroskedasticity concerns were resolved by estimating regressions using
heteroskedasticity robust (White- corrected) standard errors. It is critical to control for
cassava production given its link as a labor saving crop. A household may grow cassava
an given year in order to release labor and thus enabling it to participate in RNFE that

year creating a simultaneity bias. By using a pre-deterrnined variable on cassava

160

production before the survey season, we eliminate simultaneity bias of growing cassava
to release labor and participating in RNFE because of existence of Slack labor.

Using bivariate probit to test for simultaneity bias, we rejected the null hypothesis that
there is no evidence of participation in RNFE is linked to. growing cassava to enable Slack

labor to be released, p=0 (see Appendix C, Table C3) with a p-value of 0.08. Due to this

evidence of simultaneity bias between participating in the RNFE and the variable on
growing cassava in the survey year (2002), we therefore replaced that variable with
another one representing growing cassava prior to the survey year — cassava growing in
1997.

Results of the collinearity diagnosis revealed that using death variables split into
four categories leads to serious problems. Reducing the death variables to recent and
earlier deaths helps to partially resolve collinearity problems. The rule of thumb is that
Variance Inflation Factors (VIFS) of 10 or higher and equivalently, a tolerance of .10 or
less are reasons for concern. The variable years Since death, and death by timing by
household status had VIFS higher than 10 and were subsequently eliminated from the
analysis. Only the PA death by gender of the deceased had VIFs ranging from 1.33 to
5.12 (tolerance were above .10). The death timing impact regressions to test hypothesis
HP5 therefore used the following variables PA male recent death, PA female recent

death, PA male earlier death and PA female earlier death.

161

4.6.3 Determinants of Individual Participation in RNFE

Probit results for determinants of individual participation in rural non—farm
employment in Malawi are given in Table 4.8 for the case of death variables (base case-
no timing) while death variables with timing considerations are given in Table 4.9. Three
separate regression models were run for aggregate rural non-farm employment,
agricultural rural non-farm employment and non-agricultural rural non-farm employment.
First, we discuss the role of death variables in influencing each individual’s decision to
take participate in non-farm activities looking at both the base case and time elapsed
Since a PA death Shock and then the other controlling variables (individual, household

and location and market access).

Death Variables

Results Table 4.8 shows that contrary to hypothesis HPz, the death of a male
household head or spouse has a positive effect on an individual’s decision to participate
in the rural non-farm employment. The likelihood of participation increases by 11 percent
as individuals choose to participate pushed by a need to recover from the financial loss.
The female head or Spouse death does change the likelihood of participation. Under the
ag_RNFE, the death of a man is not significant as per our expectation while the death of a
woman is. We fail to reject the hypothesis that the death of a female head or spouse
influences individuals to pull out of the agRNFE. This supports our assertion that a
female head or Spouse death elicits a labor pull effect to meet own farm household labor

needs. The death of a female non-head or spouse increases the likelihood of individual

162

Table 4.8 Determinants of participation in rural non-farm employment by adult individuals in
rural households in Malawi, 2002 MAPAMS study. Probit Model Estimation (with PA
Death Variables)

Dependent variable: Participation (1=if individual participates; 0=otherwise)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Rural Non-Farm Agricultural Non-Agricultural
Independent Variables Employment RNFE I RNFE
Estimate Marginal Estimate Marginal Estimate Marginal
(robust std. effect (robust effect (robust effect
err.) std. err.) std. erg
Individual Characteristics
Age 0.119*** 0.032 0.063** 0.007 0.121 *** 0.020
(.022) (.027) (0.026)
Age squared -0.001*** -0.000 -.001*** -0.000 -0.001*** -0.000
(.0004) (.0001) (.0003)
Education -0.004 -0.001 -0.019 -0.002 0.006 0.001
(.020) (.027) (.021)
Female -0. 176 -0.045 -0.453 -0.053 0.183 0.030
(.146) (.187)** (.175)
Household head 0.457 0.133 0.496 0.068 0.157 0.027
(.592) (.630) (.673)
Household spouse -0.035 -0.009 0.349 0.047 -0. 195 -0.030
(.201) (.257) (.236)
Household Characteristics
Adults -0.1 l2*** —0.030 -.102*** -0.012 -0.075** -0.013
(.033) (.039) (.037)
Dependency ratio 0052 -0.014 -0.006 -0.006 -0.040 -0.007
(.060) (.008) (.066)
Female headed 0.073 0.020 0.824*** 0.157 -0.621 ** -l .070
(.235) (.300) (.269)
Age of head 0.003 0.001 0.001 0.000 0.006 0.001
(.012) (.012) (.013)
Education of head 0.038 0.010 -0.009 -0.001 0.051 0.009
(.033) (.040) (.034)
Housing Value (MK) -2.8e-06 -7.5e-07 l.le-06 1.27e-07 -4.9e-06 -8. le-07
(2.2e-06) (1.9e-06) (3.1e-06)
Land holding (ha) 0.006 0.002 -0.002 -0.000 -0.004 -0.001
(.049) (.056) (.055)
Grow cassava in 97 -0. 196 -0.049 0.1 l 1 0.014 —0.347* -0.047
(.183) (.200) (.207)
Farmer’s Club -0.275* -0.066 -0.301* -0.029 -0. 179 -0.027
membership (.164) (.210) (.178)
Location-access characteristics
Distance to bus Stop 0.003 0.001 -0.005 -0.001 0.004 0.001
(.012) (.016) (.013)
Distance to market -0.008 -0.002 0.011 0.001 -0.015 -0.003
(.010) (.012) (.013)
Kasungu ADD 0.041 0.01 1 -0.297** -0.032 0.286M 0.050
(.114) (.129) (.135)
Blantyre ADD -0.36l ** -0.090 -0.875*** -0.802 0.144 0.025
(0.147) (.208) (.161)

 

 

 

 

Source: MAPAMS 2002. *** 1% Significance, **5% significance, *10% significance

163

 

Table 4.8 (Cont’d) Determinants of participation in rural non-farm employment by adult individuals
in rural households in Malawi, 2002 MAPAMS study. Probit Model Estimation (with

 

 

 

 

 

 

 

 

 

 

 

 

 

PA Death Variables)
Rural Non-Farm Agricultural Non-Agricultural
Independent Variables Employment RNFE RNFE
Estimate Marginal Estimate Marginal Estimate Marginal
(robust effect (robust effect (robust effect
std. err.) std. err.) std. err.)
Death variables
PA male hos death 0.364** 0.110 0.267 0.037 0.209 0.039
(.175) (.210) (.197)
PA female hos death -0.156 -0.039 -0.7 l 6** -0.050 0.130 0.023
(.223) (.307) (.222)
PA male nhos death 0.189 0.054 -0.645 -0.047 0.401 ** 0.083
(.054) (.450) (.203)
PA female nhos death -0.283 0.084 0473* 0.075 -0.006 -0.001
(.191) (.274) (.209)
Constant -2.42*** --- -1.58*** --- -3.47*** ---
(.451) (.553) (.530)
Wald Chi Sq Statistic 222.85 98.51 147.79
p-value 0.000 0.000 0.000
R-squared 0.222 0.171 0.203
Log likelihood -458.04 -281.56 -352.02
N 1119 1119 1119

 

 

 

 

Source: MAPAMS 2002. *** 1% Significance, **5% Significance, *10% significance

hos = female head or spouse,

nhos = non head

participation by about 8 percent leading us to reject hypothesis HP3, which predicts a

labor pull regardless of position in the household or market under consideration. Rather,

these results support the argument that non_head or spouse members are the money-

spinners because they are free of decision-making responsibilities. Results of the

non_agRNFE regression Show that the death of a male non_head or spouse increases the

likelihood of participation by 8 percent and hence causing us to also fail to reject

hypothesis HP3. Variables on the death of male and female head or spouse were not

significant.

164

 

Death Timing Variables

Redefining death variables to capture whether it was a recent death (0-5year ago)
or an earlier death (6-13years) along with the number of years that has passed Since the
death occurred allows us to capture adjustments and resiliency of household. Table 4.9
shows these results.

Table 4.9 shows that overall, recent PA deaths (0-5yrs ago) are significant factors
in influencing participation decisions, while earlier PA deaths (those occurred from 7 —
13 years ago) are nots. The likelihood of individual participation in the RNFE increases
by 27% following the recent death of a PA man and decreases by 24% following the
death of a PA woman. Earlier death of man or woman does not affect likelihood of
participation. We therefore reject hypothesis HP4 that claims that both earlier and recent

death would influence participation decisions.

 

5 This is also confirmed in a model run separately using aggregated death variables rpam , epam and YSD.
See Appendix 4D which shows part of the results.

165

Table 4.9

Determinants of participation in rural non-farm employment by adult individuals in

rural households considering timing of PA death in Malawi, 2002 MAPAMS study.
Probit Model (with Timing of PA Death Variables)

Dependent variable: Participation (1=if individual participates; 0=otherwise)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Rural Non-Farm Agricultural Non-Agricultural
Independent Variables Employment RNFE RNFE
Estimate Marginal Estimate Marginal Estimate Marginal
(robust std. effect (robust effect (robust effect
err.) std. err.) std. err.)
Death Timing Variables
PA male recent death 0.789** 0.274 0.206 0.029 0.573* 0.143
(.319) (.306) (.304)
PA female recent -0.267** -0.242 0.431 0.071 -0.310* -0.165
death (.218) (.369) (.287)
PA male earlier death 0.094 0.025 -0.194 -0.020 0.297 0.458
(.348) (0.440) (.342)
PA female earlier 0.571 0.222 -0.447 -0.038 0.221 * 0.112
death (.482) (0.495) (.234)
Individual Characteristics
Age 0.120*** 0.032 0.069** 0.008 0.1 17*** 0.020
(.022) (.027) (0.026)
Age squared -0.001*** -0.0004 -0.001*** -0.0001 -0.001*** -0.0002
(.002) (.0003) (.0003)
Education 0003 -0.001 -0.014 -0.002 0.008 0.001
(.020) (.027) (.021)
Female -0. 169 -0.045 -0.438** -0.054 0.198 0.033
(.144) (.187) (.175)
Household head 0.432 0.125 0.449 0.064 0.198 0.035
‘ (.602) (.637) (.679)
Household spouse -0.071 -0.019 0.247 0.033 -0.182 -0.028
(.201 ) (.245) (.236)
Household Characteristics
Adults —0. 1 1 1*** -0.030 -.107*** -0.013 -0.07 l ** -0.012
(.033) (.040) (.036)
Dependency ratio -0.049 -0.013 -0.039 -0.005 -0.050 -0.008
(.060) (0.074) (.067)
Female headed 0.022 0.006 0.762** 0.146 -0.679** —0.076
(.235) (.296) (.274)
Age of head 0.002 0.001 0.001 -0.0001 0.005 0.001
(.012) (.012) (.014)
Education of head 0.032 0.009 -0.021 -0.003 0.049 0.008
(.033) (.039) (.034)

 

 

 

 

 

 

 

 

 

Source: MAPAMS 2002. *** 1% Significance, **5% significance, *10% significance

Notes:

hos = female head or spouse,

nhos = non head

166

 

Table 4.9 (Cont’d) Determinants of participation in rural non-farm employment by adult individuals
in rural households considering timing of PA death in Malawi, 2002 MAPAMS

study. Probit Model (with Timing of PA Death Variables)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Rural Non-Farm Agricultural . Non-Agricultural

Independent Variables Employment RNFE RNFE
Estimate Marginal Estimate Marginal Estimate Marginal
(robust std. effect (robust effect (robust std. effect
err.) std. err.) err.)

Housing Value (MK) -2.9e-06 -7.6e-07 6.30e-07 7.64e-08 -4.59e-06 -7.626-07
(2.20e-06) (1 .93e-06) (2.97e—06)

Land holding (ha) 0.010 0.003 -0.012 -0.002 0.004 0.001
(.048) (.056) (.055)

Grow cassava in 97 -0.207 -0.051 0.072 0.010 -0.348* -0.048
(. 178) (.199) (.201)

Farmer’s Club -0.261 -0.063 -0.290 -0.030 -0.161 -0.025

membership (.165) (.209) (.178)

Location-access characteristics

Distance to bus stop -0.001 0.0002 -0.007 -0.001 -0.001 0.0001
(.012) (.0160) (.013)

Distance to market 0006 -0.002 0.011 0.001 -0.013 -0.002
(.010) (0.012) (.012)

Kasungu ADD 0.062 0.017 -0.285** -0.032 0.304** 0.054
(.113) (.126) (.135)

Blantyre ADD -0.353** -0.088 -0.904*** -0.088 0.189 0.033
(0.144) (.210) (.156)

Constant -2.443*** --- -1.686*** --- -3.430*** ---
(.451) (.554) (.525)

Wald Chi Sq Statistic 226.21 83.13 157.61

p—value 0.000 0.000 0.000

R-Squared 0.228 0.157 0.209

Log likelihood -454.89 -286.02 -349.31

N 1 119 1119 1119

 

 

 

 

Source: MAPAMS 2002. *** 1% Significance, **5% significance, *10% significance

167

 

Focusing on disaggregated market types, we note that PA death does not
influence an individual’s decision to participate in agricultural RNFE while it does in the
case of the non-agricultural RNFE. For the non-agricultural RNFE, both recent male and
earlier female PA deaths increase the likelihood of individual participation by 14 and 11
percent respectively. However, the recent death of a PA female decreases the likelihood
of participation by 17 percent.

These results indicate that, contrary to hypothesis HP4, decision on participation
in the rural non-farm employment are sensitive to timing after death gender of deceased
and the type of market. Recent male death gives individuals a “financial push” and recent
female death a “labor pull”. In the non-agricultural RNFE, we reject the hypothesis
because PA female recent and earlier deaths have different Signs. This Shows that
decisions defer between period following the death event and after allowing time to move
on. Following the death of a female PA surviving individual’s likelihood to participate

decreases by due to a labor pull but in the long run it increases.

Other Controlling Variables

Table 4.8 shows that age has a positive and Significant effect on the likelihood of
participation, while age-squared has a negative and significant effect on participation in
all three cases. This quadratic variable reveals that the effect of age on participation is
essentially inversely U-shaped. The likelihood of participation in RNFE increases up to a
maximum age and then decreases there after. The maximum ages were computed to be
45 years for RNFEp, 35 years for agRNFEp and 46 years for non-agricultural non-farm

employment. Because the maximum age of increasing likelihood of participation is

168

lower for agricultural RNFE than for non-agricultural RNFE (ie. 35 vs 46), we conclude
that younger people gravitate more toward agricultural RNFE while older people
gravitate more toward non-agricultural non-farm employment. Corral and Reardon
(2001) found Similar results for Nicaragua.

Contrary to prior expectation, education is not a significant determinant of
participation in either agricultural and non-agricultural rural non-farm employment in
Malawi. In the literature, education has different effects. For example, Lanjouw (2001)
found education a significant determinant of probability of non-farm employment
participation as primary occupation in high productivity jobs but not significant in low
productivity jobs in El Salvador. Similarly, in Tanzania’s peri-urban areas, Lanjouw et al.
(2001) found that primary education was significant in business activities and not in non-
farm labor. It appears that the setting and type of work involved matter in the assessment
of whether or not education matters. For the level of skills needed in ganyu work
(agricultural wage) and fishing or basket making (non-agricultural 'work), it is very
plausible that education has no effect.

Individual results seem to suggest gender bias toward male participation in RNFE
exist in Malawi. Women are less likely to participate in agricultural RNF employment,
perhaps due to the likely conflict of time-sharing between on agricultural fieldwork and
domestic chores. The female variable is not a significant determinant of participation in
non-agricultural RNF employment. Glick and Sahn (1997) attributed such results to the
fact that the types of activities women become involved in are strongly influenced by the

multiple roles they play in the household.

169

Among the household variables, the total number of adults is usually associated
with increased household capacity to diversify its income generating activities, so the
larger the number of adults, the greater the probability of an individual adult’s
participation in non-farm work (Matshe and Young, 2004). However, there is also a
question of motivation or incentives. In this case, where off-farm income from
individuals is pooled for family use, the larger the number of adults, the greater the
probability of an individual not participating as he or she chooses leisure time in the
hopes that other members could participate instead. This could also be the case when the
off-farm labor market is not large enough to absorb all potential participants. Our results
show that the number of adults reduces the probability of participation in rural non-farm
employment in Malawi, both in agricultural and non-agricultural employment.
Dependency ratio is not a significant determinant of participation in RNFE.

Female-headed households normally rely on all able members of the household to
generate additional income (Sanchez, 2005). This is true for MalaWi and Southern Africa,
where there is a gender bias toward males participating in non-farm work. Our results
Show that female-headed households increase the individual’s likelihood of participating
in agricultural employment but have the opposite effect in non-agricultural employment.
Agricultural employment does not require capital outlay, while non-agricultural
employment requires some initial capital. Also cultural norms in Malawi reinforce
women’s disadvantaged position. Discriminatory inheritance practices and property
grabbing greatly disadvantage surviving widows (Munthali, 2002). Left with no assets to
sell, or access to any form of capital, financially constrained female-headed households

normally gravitate toward agricultural employment. This is true to both patrilineal and

170

matrilineal areas. Results of our focus group discussion (Appendix A3) highlighted the
plight of one woman (Mrs. Phiri — not her real name) in the matrilineal southern region,
whose husband passed away and whose relatives came and took away everything,
including iron roofing. She had her husband’s will but relatives disregarded it, all the
same. Tiessen (2005) noted that while the Malawi Parliament passed a wills and
inheritance bill in 1998, constitutional rights recognized under constitutional law often go
unrecognized under traditional laws in rural Malawi.

The age and education of the household head were not significant determinants of
individual participation in rural non-farm employment in Malawi. These results are in
contrast to what is found in literature. For instance, Sanchez (2005) found that in Bolivia,
although age of household head was not a significant determinant of individual
participation in non-agricultural self-employment, it was a Significant determinant of
wage employment in both agricultural and non-agricultural work. De J anvry and
Sadoulet (2001) found Similar results in Mexico. Evans and Ngau (1991) found
household heads that had more than 5 years or more of education more likely to have
additional sources of income outside farming in Kiriyanga district of Kenya. Age of
household head was not a significant predictor of household’s non-farm income.

Value of housing assets and total land were not significant determinants of
participation. Cassava production prior to the survey in 2002 is a significant determinant
of participation in non—agricultural employment at 10%, but with a negative marginal
effect. Contrary to expectation, prior cassava production reduces probability of non-
agRNFE. This could possibly be because the household is very labor-constrained or that

the labor released is invested in own farm operations. Prior cassava production variable is

171

insignificant for agricultural employment and overall rural non-farm employment. This
implies that perceived labor saving and labor releasing effect of cassava production are
not really tangible for individuals in the households deciding on participation.
Membership in a farmer’s club, a proxy for access to credit, is a Significant negative
determinant of individuals’ participation in overall rural non-farm employment and in
agricultural wage labor employment. So credit reduces the probability of participation.
With credit access the farm households concentrate their effort on-farm.

Among the location and market access variables, distance to bus stop and distance
to market are not Significant determinants of RNFE participation, however, regional
variables are. Geographic regions offer opportunity and incentives to diversify rural
household income (Reardon and Taylor, 1996). For Kasungu and Blantyre ADD, these
offer incentives to diversify out of agricultural RNF employment. For Kasungu, there is

a positive marginal effect of participation in non-agricultural RNF employment.

4.6.4 Determinants of Intensity Levels of Individual Participation in RNFE

This section presents results of the second hurdle analysis on the determinants of
the degree of the individual participation in the non-farm labor market, given that an
individual is participating. The truncated regression for the 255 individuals (out of 1119)
who were engaged in RNFE examines determinants of their intensity of participation in
RNFE as measured in hours per week including travel. Results of this second hurdle
model are presented in Table 4.10. We discuss death variables first and then other

controlling variables.

172

Table 4.10 Determinants of intensity of participation in rural non-farm employment by adult
individuals in rural households in Malawi, 2002 MAPAMS study. Truncated regression
( with PA Death Variables)

Dependent variable: Off-farm work time (number of hrs worked per week)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Rural Non-Farm Agricultural Non-Agricultural
Independent Variables Employment RNFE ' RNFE
Marginal Marginal Marginal
Effect Z Effect Z Effect Z
(robust Std. Statistic (robust Statistic (robust Statistic
err.) std. err.) std. err.)
Individual Characteristics
Age 0.770 1.58 0.110 0.22 0.884 1.21
(.488) (.489) (.732)
Age squared -0.007 -1.15 -.003 -0.54 -0.007 -0.84
(.006) (.006) (.0082)
Education 0.440 0.86 0.088 0.16 1.146 1.38
(.514) (.550) (.829)
Female 4.637 1.06 0.148 0.04 0.834 0.13
(4.386) (3.855) (6.323)
Household head 13.99 1.30 -2.709 -0.22 23.790 1.58
(10.69) (12.32) (15.02)
Household spouse -4.538 -1.33 6.128 0.68 -4.890 -l.33
(3.409) (8.983) (4.995)
Household Characteristics
Adults -0. 140 -0. l 8 2.120** 2.01 -0.715 -0.62
(.788) (1.055) (1.145)
Dependency ratio 1858* 1.63 2.657** 2.31 1.837 1.05
(1.139) (1.149) (1.748)
Female headed -5.969* -l.87 4.054 0.76 -6.698 -1.35
‘ (3.187) (5.314) (4.973)
Age of head -0.236 -1 .07 0.149 0.55 -0.395 -l.35
(.220) (.272) (.292)
Education of head 0.118 0.19 0.433 0.66 -0.772 -0.80
(.605) (.652) (.961)
Housing Value (MK) 1.89e-06 0.04 -.0002*** -2.63 .00001 1.32
(0.0001) (0.0001) (0.0001)
Land holding (ha) 1.667* 1.70 -0.378 -0.34 4.00*** 2.91
(.980) (1.123) (1.375)
Grow cassava in 97 -2.649 -0.90 -6.73l*** —3.05 3.458 0.61
(2.949) (2.210) (5.639)
Farmer’s Club 0.870 0.30 1.223 0.27 0.298 0.08
membership (2.890) (4.544) (3.698)
Location-access characteristics
Distance to bus stop 0.136 0.60 -0.281 -0.82 0.199 0.66
(.225) (.343) (.299)
Distance to market 01 16 -0.60 0. 162 0.62 -0.176 -0.66
(.192) (0.262) (.266)
Kasungu ADD 3.204 1.27 2.450 0.97 0.867 0.27
(.114) (2.531) (3.217)
Blantyre ADD -7.124* 1.90 7.298 1.28 3.460 0.79
(3.744) (5.703) (4.369)

 

 

 

173

 

Table 4.10 (Cont’d) Determinants of intensity of participation in rural non-farm employment by
adult individuals in rural households in Malawi, 2002 MAPAMS study. Truncated
regression (with PA Death Variables) with PA death

 

 

 

 

 

 

 

 

 

 

 

 

 

Rural Non-Farm Agricultural Non-Agricultural

Independent Variables Employment RNFE RNFE
Marginal Marginal Marginal
Effect Z Effect Z Effect Z
(robust std. Statistic (robust Statistic (robust Statistic
err.) std. err.) std. err.)

Death variables

PA male hos death -2.391 -0.78 -0.634 -0.21 -1.351 -0.26
(3.050) (3.079) (5.156)

PA female hos death -3.216 -0.89 -0.607 -0.08 -2.915 -0.58
(3.596) (7.166) (5.030)

PA male nhos death -0.729 -0.20 -5.394 -0.54 -2.035 -0.49
(3.623) (9.988) (4.128)

PA female nhos death -4.318 -1.50 -4.534** -l.95 -5.989 -1.45
(2.885) (2.321) (4.123)

Constant --- ---

LR Chi Sq Statistic 42.43 32.73 37.23

p—value 0.008 0.086 0.031

Log likelihood -975.69 -281.56 -604.18

N 255 103 154

 

 

 

 

Source: MAPAMS 2002. *** 1% Significance, **5% significance, *10% Significance
hos = female head or spouse,

Death Variables

nhos = non head

Concerning intensity of participation, hypothesis HP5 held that death variables

should not determine the actual amount of hours worked once an individual clears the

first hurdle of deciding to participate. Based on the results in Table 4.10, we reject

hypothesis HP5 because some death variables do indeed influence the level of

participation intensity in the agRNFE market. The hypothesis holds true for the non-

agRNFE but not in agRNFE market. Table 4.10 Shows that among the death variables,

only the death of a female non-head or spouse is a Significant determinant of intensity of

participation and tends to reduce agricultural RNFE by 4.5 hours weekly. This result

suggests a “labor pull” effect with a gender bias, where intensity of participation declines

174

 

for individuals in households that have had non-head or Spouse female PA death. This
variable is Significant only in agricultural RNFE, suggesting that female non-spouse
members are key players in providing labor during the growing season. The death of such
members therefore leaves a labor gap that forces individuals to cut back hours Spent off-

farm to compensate for lost labor on-farm.

Death Timing Variables

The results from death timing variables Shown in Table 4.11 also counter HP5 and
suggest that death variables do affect the intensity of RNFE labor allocation. Recent
death of PA man iS associated with a 4.9hr and 8.2hr reduction in hours worked per week
in aggregate RNFE and the agRNFE respectively. In aggregate RNFE is the death of a
PA woman associated with a reduction in 5.4 hrs per week. Earlier death of a PA man is
associated with a 17 hrs per week increase of hours worked in the non-agricultural
RNFE. No death-timing variables significantly influenced intensity of participation in
the agricultural RNFE.

The results of recent PA male death and earlier PA male death seem to suggest
that different decision behaviors are exhibited following the death of a man. An initial
response to cut back hours occurs in the short term followed by an increasing of hours
worked over the long term. Increasing the number of hours worked as the years go by
following death of a PA individual suggests that the non-agricultural RNFE may be an

important coping Strategy that supports household resiliency to a male death shock.

175

Table 4.11 Determinants of intensity of participation in rural non-farm employment by adult
individuals in rural households in Malawi, 2002 MAPAMS study. Double Hurdle

model estimation (with PA Death Timing Variables)

Dependent variable: Off-farm work time (number of hrs worked per week)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Rural Non-Farm Agricultural Non-Agricultural
Independent Variables Employment RNFE RNFE
Marginal Marginal Marginal
Effect Z Effect Z Effect Z
(robust std. Statistic (robust Std. Statistic (robust Statistic
err.) err.) Std. err.)
Death with timing variables
PA male recent death -4.951* -l.70 -2.506 -0.71 -8.193** -2.32
(2.918) (3.533) (3.535)
PA female recent -5.386* -1.74 -3.302 -1.02 -6.105 -1.45
death (2.516) (3.252) (4.218)
PA male earlier death 10.187 1.45 -0.924 -0. 14 17.03* 1.68
(7.036) (6.689) (10.15)
PA female earlier 1.621 0.25 0.843 0.10 -1.615 -0.24
death (6.491) (8.856) (6.645)
Individual Characteristics
Age 0839* 1.78 0.175 0.34 0.744 1.05
(.473) (0.519) (0.708)
Age squared - 0.008 -1.42 -0.004 -0.58 -0.006 -0.77
(0.006) (0.007) (0.008)
Education 0.386 0.73 -0.003 -0.01 1.106 1.38
(0.503) (0.553) (0.799)
Female 4.495 1.06 -0.552 -0. 14 0.601 0.10
(4.24) (3.973) (6.082)
Household head 10.647 1.03 -2.899 -0.22 189.77 1.34
(10.33) (13.20) (13.99)
Household spouse -3.3 17 -0.93 6.103 0.66 -3.127 -0.60
(3.571) (9.279) (5.214)

 

 

 

 

 

Source: MAPAMS 2002. *** 1% Significance, **5% significance, *10% significance

Notes:

hos = female head or spouse,

nhos = non head

176

 

Table 4.11 (Cont’d) Determinants of intensity of participation in rural non-farm employment by
adult individuals in rural households in Malawi, 2002 MAPAMS study. Truncated

Regression model estimation (with PA Death Timing Variables)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Rural Non-Farm Agricultural Non-Agricultural
Employment ’ RNFE RNFE
Marginal Marginal Marginal
Effect Z Effect ' Z Effect Z
(robust std. Statistic (robust Statistic (robust Statistic
err.) std. err.) std. err.)
Household Characteristics
Adults -0.088 -0.1 1 1.924* 1.80 0.920 -0.82
(.779) (1.066) (1.12)
Dependency ratio 1859* 1.69 2.41 1** 2.08 1.658 0.97
(1.098) (1.159) (1.70)
Female headed -5.686* -l.80 4.103 0.74 -6.305 -1.27
(3.153) (5.562) (4.951)
Age of head -0.150 —0.67 0.140 0.47 -0.293 -0.96
(.223) (.295) (.295)
Education of head 0.262 0.44 0.507 0.76 -0.503 -0.54
(.597) (.668) (.934)
Housing Value (MK) 4.77e-06 0.11 -.0002** -2.28 .00001 1.35
(0.0001) (0.0001) (0.0001)
Land holding (ha) 1.704* 1.77 -0.536 -0.45 4.25*** 3.19
(.965) (1.201) (1.332)
Grow cassava in 97 -2. 54 -0.87 -6.975*** -3. 17 3.493 0.62
( 2.929) (2.199) (5.669)
Farmer’s Club 0.602 0.22 1.199 0.26 -0.903 -0.27
membershiL (2.787) (4.686) (3.353)
Location-market access characteristics
Distance to bus stop 0.135 0.61 -0.231 -0.65 0.218 0.74
. (.222) (.354) (.294)
Distance to market -0.078 -0.41 0.129 0.47 -0.101 -0.38
(.191) (0.272) (.265)
Kasungu ADD 2.786 1.20 2.1 11 0.82 0.794 0.26
(2.321) (2.582) (3.090)
Blantyre ADD 7.207** 1.97 5.772 1.07 4.337 0.99
(3.664) (5.383) (4.393)
Constant —-- ---
LR Chi Sq Statistic 49.06 29.57 43.59
p-value 0.0082 0.1079 0.0059
Log likelihood -972.36 -356.71 -597.38
N 255 102 153

 

 

 

Source: MAPAMS 2002. *** 1% Significance, **5% Significance, *10% Significance

177

 

Other Controlling Variables

Pertaining to other controlling variables, Table 4.10 shows that no individual
characteristics play important roles in determining the level of participation. Once an
individual has decided to participate, only household characteristics, location market-
access and prime-age adult death influence the intensity of participation. These results
confirm the superiority of the double hurdle model over the Tobit model, by implying
that participation and intensity of participation are distinct decisions.

Among the household characteristics, the number of adults in the household was a
Significant determinant of the level of participation in agricultural employment. The
marginal effect of an adult is positive for agricultural RNFE hours. The dependency
ratio is a significant positive determinant of overall RNFE and agricultural RNFE hours
but not Significant for nonagricultural RNFE levels. AS the dependency ratio goes up,
individual participants increase the hours spent by 1.9 and 2.1 for overall RNFE and
agricultural RNFE hours respectively. These results confirm our expectation that
pressure of dependents forces prime-age adults to invest more hours raising
supplementary income.

A female-headed household is associated with reduction in the level of intensity
of participation. One possible explanation for rural Malawi is that households headed by
women decrease the level of participation of individual adults in RNFE due to the need to
share the multiple roles that the women would fill if they did not have the responsibility

of heading the household.

178

The age of household head and his/her level of education are not Si gnificant
determinants of intensity of participation. The value of housing is a significant negative
determinant of intensity level for agricultural RNFE. Individuals from well-to-do
households as measured by value of housing apparently'spent less time working in other
people’s fields doing ganyu labor. This finding corroborates observations by Bryceson
(2007) in Malawi.

The result on landholding shows that amount owned tends to increase RNFE and
non-agricultural RNFE. In this case, once an individual has decided to participate in
RNFE and non-agricultural RNFE, it seems the amount of land owned induces them to
increase hours to earn more money that can perhaps be used to purchase inputs for the
next season. Land in this regard is not a capacity variable but rather an incentive that
induces increased levels of participation.

Production of cassava prior to the year of the survey tends to reduce participation
in RNFE. This result is contrary to our expectation that for an individual who has
decided to participate, having cassava on-farrn can release time which they could invest
in agricultural RNFE, this variable is actually associated with a reduction in hours
invested off-farm. It supports earlier results that prior cassava production reduces the
likelihood of participation. The farmers’ club membership did not appear to influence
participation intensity levels.

None of the location and market access variables are Significant determinants of
intensity of participation, with the exception of Blantyre ADD for overall RNFE. This

variable is Significant at 10% with a large positive marginal effect. This result suggests

179

that for an individual that has decided to participate in RNFE, Blantyre ADD and its large

city has the geographic setting to pull them into rural non-farm employment.

4.6.5 Summary Discussion Comparing the Role of Death Variables in Influencing

Decisions on Participation and the Intensity Levels in RNFE Markets.

Death variables clearly have major effects on RNFE in contemporary Malawi,
where HIV/AIDS has made prime age adult death endemic. The gender and position of
the deceased PA individual in the household are critical factors in influencing
participation and intensity decisions. We found gender differences where the death of a
woman did not affect the likelihood of participation in the higher paying non-
agricultural RNFE, while it did in the agricultural RNFE - reducing likelihood of
participation. While we hypothesized that the death of a PA head or spouse would
result in a leadership vacuum that pulls individuals back on the farm, our results showed
that a financial loss is much more felt than a leadership vacuum. Indeed, prime age
mortality is always associated with a reduction in hours worked, especially in
agricultural RNFE.

Timing of the death shows that recent deaths of 0 — 5 years ago are Significant
determinants of RNFE while earlier deaths of 6 — 13 years are only important in the case
of the non-agricultural RNFE. The death of a PA man triggers different initial and post-
adjustment responses, where individuals cut back hours in the Short term and then
increase them over time, suggesting this could be a strategy to shore up household

resilience to the death shock.

180

The emerging story is that PA death in the era of HIV/AIDS can affect an
individual adult’s decision to participate in rural non-farm employment, depending on
the gender and household position of the deceased. A “labor pull” effect occurs with the
death of PA female heads or spouses, as home care labor demand causes retraction of
labor from the non-farm to focus on the farm household. A “financial push” effect
occurs with the death of a PA man or to some extent non-head or spouse women who
are “money-spinners”. Decisions to participate and the level of intensity of
participation in off-farm employment differ by time Since death occurred and by the

type of RNFE.

4.7 Conclusions

We set out to examine how death stricken households respond to RNFE labor
markets and how their responses changes over an adjustment period Since the initial
Shocks. We modeled the determinants to individual adult participation and intensity
levels of participation (hours/week) of individuals in agRNFE, non-agRNFE and
aggregate RNFE using cross-sectional data from rural Malawi. The adult death status of
afflicted households was captured by variables identifying the gender of the deceased, his
or her position in the household, (ie. whether a head or spouse or non-head or spouse),
and the number of years passed since the death shock occurred.

Our results show that prime age adult death has been a significant determinant of
participation in RNFE in Malawi Since 1989—90. First, the gender and position of the
deceased matters as far individual’s decisions to participation is concerned. Death of a

PA adult male head or spouse had a financial push effect whereby individuals were

181

pushed towards participating in RNFE to recover lost financial gains. The death of a PA
adult female head or spouse, on the other hand, was associated with a “labor pull” effect
where individuals pulled back out of deciding to participate in the labor market.

Second, death of a PA woman, regardless of status, influenced participation
decisions only in the agRNFE indicating there could be a gender barrier to entry in the
higher payoff, non-agRNFE market or that women prefer the agRNFE. More research
would be needed to search this out.

Third, the death of a man who was neither a household head nor a spouse
Significantly influenced individual participation in non-agRNFE by inducing a “financial
push” into the non-farm employment.

Among adults working in rural non-farm employment, only the prime age adult
death of a non-household head or Spouse affects the number of hours Spent in agricultural
RNFE. Such a death is associated with a decline in hours Spent on agricultural RNFE
reinforcing perhaps, the gender division of labor where all house chores and orphan care
are considered woman’s work.

As for the timing of death variables, our study Showed that off-farm participation
decisions differ by time elapsed Since death. Individuals are more likely to participate in
the 0-5 years after death and less likely to over the long term (6-l3years following a PA
death). For those who participate, particularly in non-agricultural RNFE, the initial level
of intensity of participation drops in the days following the death shock but over time, the
hours worked increase, implying that engaging in RNFE is important for strengthening

afflicted households’ resiliency to death shocks.

182

Given the importance of income obtained from rural non-farm employment in
alleviating food insecurity, rural poverty and improving agricultural productivity and
rural livelihoods, policies aimed at addressing cultural and economic barriers to entry in
non-agricultural RNFE are needed. Such policies need to be put in place along with
broader rural development ones that ensure the labor market grows as well. Such growth
can be put in place through policies that promote upstream and downstream linkages such
as agricultural input and product processing, small-scale rural manufacturing other

support services.

Limitation of this study and areas of further research

This study was based on cross-sectional data from a sample taken from high
maize producing areas of Malawi in 2002. The results may apply only to higher agro-
potential regions and effectively lack assessment of the dynamic effects of time sensitive
death variables that are better captured using panel data. Satriawan and Swinton (2007)
while studying the impact of human capital variables on rural non-farm income found
that education variables that were not significant using cross-sectional analysis, were
significant using panel data that enabled them to control for household fixed effects. A
further study using panel data in Malawi could rule this out and is needed to shed more
light of death impact on RNFE.

Another critical issue is access to RNFE. In the agricultural RNFE, the jobs are
ubiquitous and any PA adult who is physically able to work and willing to do so can

work. With non-agricultural RNFE, if no opportunities exist, an individual may not have

183

a platform from which to choose to participate. In other words, the household without the
opportunity of access cannot technically make a decision to participate in a non-existent
RNFE. There is need to inventory RNFE opportunities in Malawi from which plans can
be made to spur rural development programs that give opportunity to individual S in rural
Malawi.

Gender inequality remains a major issue in rural Malawi. Cultural norms reinforce
women’s disadvantaged position. Discriminatory inheritance practices and property
grabbing disadvantage-surviving widows remains a challenge (Munthali, 2002). Tiessen
noted that while the Malawi Parliament passed a wills and inheritance bill in 1998,
constitutional rights recognized under constitutional law often go unrecognized under
traditional laws in Malawi. Our study found gender differences where the death of a
woman did not affect the likelihood of participation in the higher payoff non-agricultural
RNFE, but did reduce the likelihood of participation in agricultural RNFE. Given the fact
that our study looked at choices and not at opportunities, it could not speak with authority
on whether the gender pattern observed is because of women’s preferences and
opportunities or barriers that bar women from participating in the better paying non-
agricultural market. We therefore recommend that further studies be undertaken to
examine gender preferences, opportunities and barriers to accessing rural non-farm

employment in Malawi.

184

References

Anglewicz, P., S.B. Assche, P. Fleming, A.V. Assche, C. vande Ruit. (2005) “The Impact
of HIV/AIDS on Intra-household Time Allocation in Rural Malawi” . A paper
presented at the IUSSP Seminar on “Interactions between Poverty and
HIV/AIDS” 12-13 December 2005. Cape TownSouth Africa.
httpzllwww.cssr.uct.ac.za/saldru_poverty_hivaids_papers.html

Barnett. T. and P. Blaikie. (1992) AIDS in Africa: Its Present and Future Impacts. New
York: Guilford Press. USA

Barrett, C.B., T. Reardon, P. Webb. (2001) “Nonfarm income diversification and
household livelihood strategies in rural Africa: concepts, dynamics, and policy
implications.” Food Policy 262315-331

Beegle, K. (2003). “Labor Effects of Adult Mortality in Tanzanian Households.” World
Bank Policy Research Working Paper No. 3062. Washington, DC. : The World
Bank.

Berdegue, J ., E. Ramirez, T. Reardon, G. Escobar (2001). “Rural Non-farm Employment
and Incomes in Chile.” World Development 29 (3):411-425.

Bryceson, D. F (2007). “Ganyu casual labor, famine and HIV/AIDS in rural Malawi:
causality and casualty. Journal of Morden African Studies, 44, 2 pp 173-202.

Burton, M., R. Dorsett, T. Young, (2000). “An investigation of the increasing prevalence
of non-purchase of meat by British households. Applied Economics 32: 1985-
1991.

Corral, L. and T. Reardon (2001). “Rural Non-farm Incomes in Nicaragua.” World
Development 29(3):427-442.

Cragg, J. G. (1971). “Some Statistical Modles for Limited Dependent Variables with
Application to the Demand for Durable Goods.” Econometrica 39: 829-844.

da Silva, J. G. and E. Del Grossi (2001). “Rural Non-farm Employment and Incomes In
Brazil: Patterns and Evolution.” World Development 29(3): 443—453.

de Janvry, A. and E. Sadoulet (2001), “Income Strategies Among Rural Households in
Mexico: The Role of Off-farm Activities.” World Development 29 (3): 467-480.

De Walque, Damien (2006) “Who gets AIDS and How? The Determinants of HIV

Infection and Sexual Behaviors in Burkina Faso, Cameroon, Ghana, Kenya and
Tanzania.” World Bank Policy Research Working Paper No. 3844.

185

Dirven, M. (2004). “El empleo rural no agricola y la diversidad rural de America Latina”.
Revista de la CEPAL: 49-69. Quoted in Sanchez 2005.

Ellis F. and E. Allison (2004), “Livelihood diversification and Natural resources access.”
Overseas Development Group. University of East Anglia, UK. LSP Working
Paper 9. A paper prepared for Livelihood Support Programs, Food and
Agriculture Organization, UN. ‘
ftp://ftp.fao.org/docrep/fao/006/ad689e/ad689e00.pdf

Escobal, J. (2001). “The Determinants of Non-farm Income Diversification in Rural
Peru.” World Development 29(3):497-508.

Evans, HE. and P. Ngau (1991), “Rural-Urban Relations, Household Income
Diversification and Agricultural Productivity’. Development and Change, 22,
519-545.

Ferreira F.H.G. and P. Lanjouw (2001). “Rural Non-farm Activities and Poverty in the
Brazilian Northeast”. World Development 29 (3): 509-528.

Garcia, J. and J .M. Labeaga (1996). Alternative approaches to modeling zero
expenditure: an application to Spanish demand for tobacco. Oxford Bull. Econ.
Stat. 58, 489-506.

Glick, P. and DE. Sahn (1997). “Gender and education Impacts on Employment and
Earnings in West Africa: Evidence from Guinea.” Economic Development and
Cultural Change: 794-823.

Greene, w., (2002). Econometric Analysis, 5th Edition. Macmillian, New York, USA

Heckman, J ., (1979). “Sample Selection bias as a specification error.” Econometrica 47:
153-161.

Huffman, W.E. (1980). “Farm and Off-farm Work Decisions: The Role of Human
Capital”. The Review of Economics and Statistics, 62 (1): 14-23.

Huffman, W.E. and M. Lange. (1989) “Off-farm work Decisions of Husband and wives”.
The Review of Economics and Statistics, 71 (1): 471-80

Hussein, K. and J. Nelson (1998). “Sustainable livelihoods and livelihood
diversification”. Brighton. IDS Working Paper 69, 32p. Africa/Asia. No. IDS

website.

Jones, A. M., (1989). “A double hurdle model of cigarette consumption.” Applied
Econometrics. 4: 23-39.

186

Kerr, R. B. (2005). “Informal labor and Social Relations in Northern Malawi: The
theoretical Challenges and Implications of Ganyu labor for food security” Rural
Sociology: 70(2):

Lanjouw, J ., and P. Lanjouw (2001). “The rural non-farm sector: issues and evidence
from developing countries.” Agricultural Economics 26: 1-23.

Lanjouw, P. (2001). “Non-farm Employment and Poverty in Rural El Salvador.” World
Development 29(3):529-547.

Lanjouw, P., J. Quizon, and R. Sparrow (2001). “Non-agricultural earnings in peri-urban
areas of Tanzania: evidence from household survey data.” Food Policy 26: 385-
403.

Matshe, I. and T. Young (2004). “Off-farm labour allocation decisions in small-scale
rural households in Zimbabwe.” Agricultural Economics 30(3): 175-186

Munthali, A. (2002). “Adaptive Strategies and Coping Mechanism of Families and
Communities Affected by HIV/AIDS in Malawi.” A draft paper prepared for the
UNRISD project of HIV/AIDS and Development”. March 2002.

Reardon, T. (1997). “Using evidence of household income diversification to inform study
study of rural nonfarrn labor market in Africa. World Development, 25(5), 735-
748.

Reardon, T. (1988). “Rural Non-farm Income in Developing Countries.” The State of
Food and Agriculture, Rome. ‘

Reardon, T., J. Berdegue, and G. Escobar (2001). “Rural Nonfarm Emploment and
Incomes in Latin America: Overview and Policy Implications.” World
Development 29(3):395-409. '

Reardon, T. and J .E. Taylor (1996). Agroclimatic Shock income inequality and poverty:
Evidence from Burkina Faso. World Development, 24(5): 901-914.

Reardon, T., J.E. Taylor, K. Starnoulis, P. Lanjouw, A. Balisacan. (2000) “Effects of
Nonfarm Employment on Rural Inequality in Developing Countries: An
Investment Perspective”. Journal of Agricultural Economics, 51(2):

Ruben R. and M. Van Den Berg (2001). “Non-farm Employment and Poverty
Alleviation of Rural Farrrr Households in Honduras.” World Development 29 (3):
549-560.

Rugalema, G. “AIDS and Rural livelihoods: A case study of smallholder households of
Bukoba District, Tanzania.” PhD Thesis. Institute of Social Studies, The Hague.

187

Tockle, J .G. and WE. Huffman (1991). “Local Economic Conditions and Wage Labor
Decisions of Farm and Rural Nonfarm Couples”. American Journal of
Agricultural Economics, 73, (3):652-670.

Sanchez, V. (2005). “The Determinants of Rural Non-farm Employment and Incomes in
Bolivia”. MSc. Thesis, Michigan State University, East Lansing MI, USA

Satriawan, E. and SM. Swinton. (2007) “Does human capital raise farm or nonfarm
earning more? New insight from a Rural Pakistan household panel.” Agricultural
Economics 36: 323-330.

Strauss, J ., and D. Thomas. (1988) “Health, nutrition and economic development.”
Journal of Economic Literature 82:766-817.

Sadoulet, E., and A. Janvry (1995). Quantitative Development Policy Analysis,
Baltimore: Johns Hopkins University Press.

Tiessen, Rebecca. (2005). “Mainstreaming Gender in HIV/AIDS Programs: Ongoing
Challenges and New Opportunities in Malawi”. Journal of lntemational Women’s
Studies. 7(1): 8-25

World Bank (1999). Confronting AIDS: Public Priorities in a Global Epidemic. Revised
edition. New York: Oxford University Press.

Wooldridge, J .M. Introductory Econometrics: A Modern Approach, South-Western
College Publishing, and Cincinnati, 2002

Yunez-Naude, A., and E. Taylor (2001). “The Determinants of Non-farm Activities and
Income of Rural Households in Mexico, with Emphasis on Education.” World
Development. 29(3):561-572.

Zwick K. and D. Smith (2001) “Access to Rural Non-farm Livelihoods: Report of

Preliminary Field Work in Rakai District, Uganda”. Natural Resources Institute,
Report No:2592. Rural Non-Farm Economy. March 2001.

188

CHAPTER 5

SUMMARY OF FINDINGS AND POLICY IMPLICATIONS

5.1 Background

Mortality of sexually active andhighly productive prime-age adults has increased
significantly with the advent and spread of the Human Immunodeficiency Virus (HIV),
the virus that causes acquired immune deficiency syndrome (AIDS). Originally viewed
as a health sector problem, AIDS has become a pandemic that is recognized as a multi-
sectoral long-term development challenge around the world and in particular Sub-
Saharan Africa. In Malawi, a largely agrarian economy, the national prevalence rate is
15 percent presents a daunting challenge to not only rural development but also the
welfare of rural households.

Since the advent of the pandemic, researchers have grappled to understand the
impacts of the resulting adult morbidity and mortality, with a view to developing
measures to mitigate the negative effects of prime-age adult death. Earlier cross sectional
surveys and case studies indicated that AIDS could adversely affect labor supply and
remittances to agriculture, causing a possible shift from labor and cash intensive mostly
cash crops to low labor intensive food and subsistence crops (Barnett, et. al, 1995).
However, these studies lacked counterfactual analysis (Mather, et al, 2005). Rigorous
impact studies that used panel data methodology later Showed that at the household level,
an adult death does not necessary equate to loss of labor as other family members often
joined the household to take responsibility (Ainsworth and Semali 1998, Beegle 2003),

thus disputing earlier recommendations that advocated labor saving technologies. Indeed,

189

Jayne et al. (2005) caution that labor saving interventions may be appropriate for certain
types of households and regions, but not across the board.

Despite the acknowledgement that AIDS is a long-term pandemic; there have
been few (but increasing) Studies that try to estimate leng-term impacts (Beegle, et al
2006, Drinkwater, et. al, 2006, Chapoto and Jayne, 2006). The majority of these studies
focus on short to medium term (4-6 year) periods, due to data scarcity. The long-term
effects of prime age adult death have been little examined and are not well understood.
Three major research questions arise. First, what is the impact on agricultural production
of prime-age adult mortality (PAM) induced losses of human and financial capital?
Second, what is the impact of PAM on afflicted households’ labor participation in the
rural non-farm sector? Third, do these impacts differ in the short term, versus medium
and long term? If so, what policies can be put in place to help ameliorate the impacts?
Answers to these questions generate important insights about intervention programs to
counter household level effects of afflicted households.

The purpose of this study was to investigate the impact of prime age adult death
on agricultural production and non-farm employment among rural households in Malawi.
In particular the study addressed three broad objectives: 1) to assess the impact of prime-
age adult mortality on agricultural production and relative effects of loss of human and
financial capital, ii) to evaluate how death stricken households respond to rural non-farm
employment labor markets and how the response changes with time Since death, and iii)
to recommend policy responses that ameliorate the negative impacts of prime-age death

on households.

190

The study focused on 351 households from Blantyre, Kasungu and Mzuzu
Agricultural Development Divisions of Malawi’s main maize producing areas. A thirteen
year panel data-set constructed from a 2002 re-study of a 1990 Maize Variety and
Technology Adoption Survey is used to answer questions on farm production impacts,
while the recent 2002 cross-sectional data from Malawi Agricultural Production and
Adult Mortality Study (MAPAMS) survey is used to answer questions on rural non-farm
employment participation. The research findings are summarized in section 5.2 below

while policy recommendation and future research needs are presented in section 5.3.

5.2 Summary of Findings

We analyzed the longitudinal data set using an ordinary least squares, fixed
effects regression to assess the impact of prime age adult death on agricultural production
as measured by total area planted as well as area planted to maize and non-maize crops
separately. For the cross-sectional data, we used a double hurdle model composed of a
first-stage probit model and followed by a truncated regression on non-zero observations
in order to assess the impact of adult death on participation and intensity of participation
in rural non-farm employment. Summaries of the results are given below, beginning with
farm production impacts and then prime age adult mortality impacts on rural non-farm

employment.

191

5.2.1 The Impact of Prime age Mortality on Agricultural Production

Overall death impacts

The study showed that prime age adult mortality does affect agriculture through
reduction of total area planted. Total area planted fell by 0.28 hectares on average among
households impacted by PA mortality irrespective of the gender and or household
membership position of the deceased. The effect of prime age death was sensitive to the
gender and position of the deceased. Households that suffered the death of a male
household head or spouse of a household head reduced their planted area by nearly a half
a hectare (0.49), while those experiencing female head or spouse lost about three eighths
of a hectare (0.37ha). The death of household members who were not heads or spouses
of heads did not impact total area planted.

Disaggregating between area planted to maize and area planted to other crops, the
death of a PA man who was ether a household head or spouse led to reduced maize area
planted by slightly over a third of a hectare. More interesting, the death of the elderly
(60-79 years of age) significantly reduced area planted to maize by a quarter of a hectare.
We attribute this to the fact that these farmers were in their prime years during the peak
of former president Kamuzu Banda’s “Chimanga Ndimoyo ” (Maize is life) policy when
agricultural extension was especially well funded and effective. Hence, these elders
gained a wealth of experience and knowledge, which was lost with their death. Chapoto
(2006) found death of the elderly man 60 years old and above reduced cereal cultivated

area in 1999/2000 by 20 percent in rural Zambia and attributed it to contribution to crop

192

production through capital resources for renting like animal draft power and other inputs
but not their physical labor input.

Prime-age mortality influenced area planted to non-maize crops as well. The area
planted to non-maize crops declined by 0.16 hectares 011 average among afflicted

households. Unlike the case of maize, elderly deaths were not Significant.

Death impacts over time

Using binary variables to indicate death in the last 0-3 years, 4-6 years, and 7-13
years, we were able to analyze short-term, medium-tem and long-term impacts of prime
age mortality respectively. Results showed PA mortality had a short-terrn (0-3 years)
effect reducing total area planted by 0.32 hectares on average. Over the long term,
however, the reduction was negligible. The largest decline in total area planted of 1.2 ha
occurred following the death of a PA male household head or spouse. The female PA
deaths did not cause any Significant decline.

When planted area was disaggregated by maize versus other crops, maize area
was particularly hard hit. When a PA male head or Spouse died, area planted to maize
declined by 0.8 hectares on average in the short term period of 0-3 years following a PA
death, shifting from 0.8 hectares to 0.53 hectares in the medium term of 4 to 6 years
following a PA death, a minor recovery. Area planted to crops other than maize was less
affected by prime age mortality, being only significant in the 0-3 year period after death,

with non-maize cultivated area reduced by 0.23 hectares.

193

5.2.2 The Impact of Prime-age Mortality on Rural Non-farm Employment

Overall death impacts

Our results show that prime age adult death Significantly affected participation in
rural non-farm employment (RNFE) in Malawi since 1989-90, and the effect depended
on gender and household position of the deceased. Death of a PA male head of household
or spouse had a “financial push” effect whereby surviving household members were
pushed towards participating in RNFE to recover lost financial gains. The death of a PA
female head or Spouse, on the other hand, was associated with a “labor pull” effect where
individuals withdrew from the rural non-farm labor market.

Second, the death of a PA woman, regardless of status, influenced participation in
the agricultural RNFE, indicating that there could be a gender barrier to entry in the non-
agricultural RNFE market or that women prefer agricultural RNFE. More research is
needed discern the difference.

Third, the death of a man who is neither household head nor a Spouse induced a
significant “financial push” toward non-agricultural RNFE among surviving household
members.

Among adults engaged in RNFE, only the death of a PA non-household head or
spouse affects the number of hours Spent on agricultural RNFE. Such a death is
associated with a decline in hours Spent on agricultural RNFE reinforcing perhaps, the
gender division of labor where all house chores and orphan care are associated with

woman’s work.

194

Death impacts over time

Timing of the death Shows that recent deaths within 0-5 years Significantly affect
RNFE participation while earlier deaths (6 - 13 years past) affect only non-agricultural
RNFE. A male PAM triggers different initial and post-adjustment responses. In the short
term, household members cut back hours, but they then increase them over time,
suggesting that this could be a resiliency strategy: Cut back travel costs and capital outlay
immediately, consolidate production on farm to minimize household food production
loss, and then over the years expand income from non-employment to regain pre-death

shock welfare level.

5.3 Policy Implications and Future Research Needs

This study Shows that PA mortality does affect the economic welfare of farm
households in Malawi via both agricultural production and off farm rural non-farm

employment participation.
Farm production policy implications

On the farm, this study has generated the following findings. First, prime age
adult mortality in Malawi reduces agricultural production through loss of area planted.
Second, the gender and household position of the deceased are important in
distinguishing the magnitude of effect on area planted. The impact of the death of male

prime age individuals who were household heads was greater than that of their female

195

counterparts. Third, non-maize area was more sensitive to prime age death shock than
maize area in the short term where a non-head or spouse death occured. Fourth, the loss
of farming experience associated with elderly deaths appears to be more important
associated with PA adult death. Fifth, the PA death effects on crop production are felt in
the short-to-medium term (0 -6 years after death), and afflicted households respond by
decreasing area planted to high capital and labor input requiring non-maize crops in the
Short term and then reducing maize area in the medium term.

From these findings, four observations are relevant for policy recommendations:
1) the decline in maize area planted may threaten food household food security among
afflicted households. ii) The strong negative impacts of deaths by male PA household
heads or Spouses imply serious vulnerability of widows, suggesting whom to target for
intervention. iii) Death impacts were more severe in the period 0-6 years after a PA
death, suggesting when to target intervention. iv) Non-maize crop production was more
sensitive to tinting of death than maize area, suggesting which crops to target
intervention.

In view of the above issues, we recommend that policy makers recognize and use
the gender and position of the household of the deceased as markers of identifying
households that need intervention. Widow-headed households would have priority in
order avert household food insecurity. Interventions Should be targeted toward short-term
(0—3 years) and medium term (4-6 years) after PAM. Short-term intervention with direct
food aid is recommended along side with a long-term strategy. Subsidized seed and

fertilizer and labor hire to counter reduced non-maize crop production due to reduced

196

planted area in the short term is a must. This will ensure households’ maize cropping

programs will not decline in the medium term.

Non-farm employment policy implications

PA death in the era of HIV/AIDS also affects the decisions of surviving adults on
participation in rural non-farm employment, depending on the gender and household
position of the deceased. A “labor pull” effect occurs with the death of PA female heads
or spouses as home care labor demand causes retraction of labor from the non-farm
sector to focus on the farm household. A “financial push” effect occurs with death of a
PA male as well as to some extent female non-head or spouses who are “money-
Spinners”. Decisions to participate and the level of intensity of participation in off-farm
employment differ by time since death occurred and by the type of RN FE.

Given the importance of rural non-farm employment in Malawi, we recommend a
comprehensive approach to promoting rural non-farm employment as a means of
supporting rural household incomes. Gender sensitive small projects should be
launched to help afflicted households to do non-agricultural work on-farm if time

constraints do not allow them to travel to off-farm.

Complementarities between farm and. non-farm intervention strategies

Certain farrrr and non-farm intervention strategies are complementary. For

instance, provision of credit under non-farm employment Si gnificantly reduced the

197

likelihood of participation. So our recommendation to give credit to farms in the short
term would increase labor and land productivity, thus potentially averting household food
insecurity. Given the time constraints of households afflicted by PA male death
households that withdraw from non-farm employmentcan use credit to grow higher
valued non-cash crops from whose sale they can pay off credit and use extra revenue to
start home based non-agricultural RNFE. Complementarities such as this one will
ensure minimum disruption of maize crop production, profitable use of retracted labor in

the interim period, and Shortening of the recovery period for non-agricultural RNFE.

Further Research

We recommend two issues for further studies. First, there is a need to further
understand the role played by the elderly in agriculture in Malawi. We speculated that
perhaps their death causes knowledge loss. Understanding why the variable is significant
can be key to designing policies on generational knowledge transfer. Given PA
individuals are dying, perhaps there is a need to bring the under 15 together with the over
59 to ensure indigenous and accrued farming knowledge is not lost. Second, we
recommend that further studies be undertaken to examine gender preferences,
opportunities and barriers to accessing rural non-farm employment market in Malawi.
Our study observed choices made by women but we could not say whether those choices
are a result of preference, limited opportunities, or cultural and economic barriers. These
further studies could Shed more light to ameliorate the negative economic impacts of

adult death on surviving household members.

198

References

Ainsworth, M., and I. Semali. (1998). “Who is most likely to die of AIDS? Socio-
economic Correlates of Adult Deaths in Kagera Region, Tanzania”. In
Confronting AIDS: Evidence from the Developing World, (ed). M. Ainsworth, L.
Fransen, and M. Over 95-110. Washington, DC: World Bank.

Barnett, T., J. Tumushabe, G. Bantebya, R. Sebuliba, J. Ngasongwa, D. Kapinda, M.
Ndileke, M. Drinkwater, G. Mitti, M. Haslwimmer, M. (1995). “Final Report:
The Social and economic Impact of HIV/AIDS on farming systems and
livelihoods in rural Africa: Some experience and lessons from Uganda, Tanzania
and Zambia.” Journal of lntemational Development 7: 163-176.

Beegle, K, (2003), “Labor Effects of Adult Mortality in Tanzanian Households”, World
Bank Policy Research Working Paper no. 3062, World Bank, Washington, DC.

Beegle, K., J. De Weerdt, and S. Dercon. (2006). “Adult Mortality and Economic Growth
in the Age of HIV/AIDS. Draft paper. World Bank, Washington, DC.

Chapoto, A. (2006). “The Impact of AIDS-related Prime-Age Mortality on Rural Farm
Households: Panel Survey Evidence from Zambia”. PhD Dissertation.
Department of Agricultural Economics, Michigan State University, East Lansing,
MI. USA

Chapoto, A. and T. Jayne (2006). “Impact of AIDS-Related Mortality on Rural Farm
Households in Zambia: Implications for Poverty Reduction Strategies.” Mimeo.
Michigan State University, East Lansing, MI. USA.

Drinkwater, M. M. McEwan and F. Samuels. (2006). “The Effect of HIV/AIDS on
Agricultural Production Systems in Zambia: A Restudy 1992-2005. Analytical
Report. www.ifpri.org/renewal/pdt/Zambia_AR.pd1

Jayne, T. S., M. Villarreal, P. Pingali, and G.Hemrich. (2005). “HIV/AIDS and the
agricultural sector in Eastern and Southern Africa: Anticipated Consequences.” A

Paper presented at the lntemational Conference on HIV/AIDS and Food and
Nutrition Security, Durban, 14-16 April 2005.

Mather, D., C. Donovan, T. Jayne and M. Weber. (2005). “Using Empirical Information
in the Ea of HIV/AIDS to Inform Mitigation and Rural Development Strategies:

Selected Results from African Country Studies”. American Journal of
Agricultural Economics 87(5): 1289-1297.

199

APPENDICES

200

APPENDIX A

Survey Instruments and Focus Group Report

201

Appendix A1: Focus Group Discussion Checklist

MALAWI AGRICULTURAL PRODUCTIVITY AND ADULT MORTALITY
SURVEY (MAPAMS)

FOCUS GROUP DISCUSSION (F GD) CHECKLIST

MAIN GROUP

1. Wealth Ranking

The group should help us determine wealth categories. What are the characteristics of
household they consider to be:

Rich:
Better Off
Poor:

Very Poor:

2. Marriage System

What is the dominant marriage system in this community? Matrilocal or Patrilocal?
Is it possible to have daughters in the same family married under different systems?
How much bride price is paid in Marriage?

3. Livelihood Activities

What are the livelihoods activities engaged by households in this community?

Rank them in order of importance.

4. Constraints to farnring

What are the main constraints to farming in this area?

202

5. Causes of chronic illness and deaths and losses associated with them
What are the main causes of Sickness and death in this community?

How are chronic illness and deaths affecting the households? i.e. what are the losses they
face?

How are households coping?

What help is there in the community, extended families, NGOS, Government etc?

SMALL GROUPS

5. Asset Depletion
What things are considered key assets in a household?

What is happening to assets when a man is (1) Sick and (2) when he dies?
What is happening to assets when a woman is (l) sick and (2) when She dies?

Access to assets post death:
What normally happens to the remaining adults and children when somebody dies?

How has this changed patterns of land access, inheritance practices, etc.) as well as the
potential impact on survivors?

What are the coping mechanisms that are available (i.e. what type of support is offered by
relatives from town, in the village, the community, any program interventions)?

Is wife inheritance practiced here?
When assets are liquidated during illness, funeral and post funeral ceremonies are these

assets replaced afterwards to get to the same level? i.e. does the illness and funeral shock
lead to a permanent or temporary decline in asset position?

203

7. Knowledge of HIV/AIDS

What is HIV/AIDS? How did they come to know about it?

What are the common symptoms of HIV /AIDS sufferers?

How can you tell whether one is HIV positive?

How does one get HIV/AIDS?

What are the preventative measures?

Are people practicing these measures?

Is the practice of multiple sexual partners outside marriage common here?
What can you as man or women do to protect yourselves and families?
What can the community do?

Have you talked about HIV/AIDS in your family?

Which are the most places one can contract AIDS around here and why?

8. Child Headed Families

Are there child-headed households in your communities?
How are they coping? What assistance do they get from the community?
Where do they learn farming methods?

9. Frequency of funerals

What has been the frequency of funerals in your community?

204

MALAWI AGRICULTURAL PRODUCTIVITY AND ADULT MORTALITY
SURVEY (MAPAMS)

FORMAL QUESTIONNAIRES

205

Appendix A2: MAPAMS Survey Instruments

MALAWI AGRICULTURAL PRODUCTIVITY AND ADULT MORTALITY SURVEY
Michigan State University/APRU-Bunda College
CIMMYT/MOA Panel Study Follow Up

November 2001 — October 2002

HOUSEHOLD ROSTER BOOK: MAPAMS —1

206

INDENTIFICATION OF HOUSEHOLD

 

Homestead Identification Remarks:

 

 

 

NB: Enumerator Notes: Please complete the appropriate shaded areas before the survey begins

Definitions:
I . Household defines a group of people who eat food prepared from the same kitchen/fire
place/nkhokwe
2. A person qualifies to be considered a household member if'he/she stays for 6weeks or more during
the farming season or for 8 weeks or more during the oflifarming season
3. The farming season is the period from October to June/July
4. 0n section HR—D Death History section ask about all details for funerals in household up to

HRD26, from HRD27 to HRD80 ask only for deceased who died at the age I5 years and above

207

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.....§.5nnn Eadie
8:89—30: .8 gun—omum
Bu_.=_..o.._m=ae:omun
....... eschew an u use b.2253 E23982. 2.5 2808
.eaeeaeone .85. 5285"” 4 BE... 52 .85 swine momzuk .5293 2.5 ensure
5.53 usemne .83 ”is?“ 838.3 E .Eeane ...> ...m -> Benn 3:838: .e Eugene
53...... 83......» .34 3.33 352;. been"... Nu 353.885 2 Em i E sane .. Em 1. Ewan aeuaoeemum
Ease 8.2%... 3.3564 :5. S esea... Meagan. .225. .85. 5: he 3.5: uses; cesium mzozfl 9235.34“ cam".
2.50 nun Sauteed: 3.50 can meta—r539 2.60 E... coir—omen 350 E... gun—€85
3% £32 Home: seam .352 “was... .23 .2885 85...: ”:3... e8: 2 . . . .. ”2....
m
a
.
we
a
e
m
.4
e53.. e.5, .. eeeu 3.5 2.5 .e2 .e2
.e2 a. mum .2".
25.2 MBLEoE w..osEoE 3.2m 3:22 .95». 5:3..an can: 2 9.5.5.533. xom om<
29:. e5:— mfi... 45:. 2;... ~55 .<~.I..l

 

 

 

 

BBEoE 29.8.5...

3:32.582 22.8.5: 25.3... 5:. 2.3

.\3e§ @223 2%. unnaturaange 8.3.:
E 35.3: 3.5.8.5.: \e 3:5: .3: 5.35%.? S .... ..BS is \e .5: 2: $8.32.»: . a: 9 Bets .2: .32 ~53. .1353: a: ham 5.85.8 :9. 3.: E. 323: ha: ESE»... 35.5% 3: ...e wick

2.31.: 2.53:. .825: ES Tau. 553.95% fine: 8 $5,555.: 68.. can» Statezango 35.85%»... EEK {SPEEEWN moi—eon— 805 .3 8:5: 2: 0.53 :33 .05: $5 .5 E395

203 Be... .3. .2. 35.8... sex 32:85 a... E can .320: 35.62 3.2 E 88:95.. 3332... S. E 3.8.5....8 20:820.. fl... . 3-93. 5

...oEo viaduvo—Socx .8 can: Eon—83°: 3:99:58: 395.

4:595:sz QQOEHWDOI waxy—43:3 35:.— ZO—HUm—m

208

amt: $55 $55. .3558 2.2: .35..qu 39.5.... as 59.5.»on .8: .85 *

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.555...
1:50 350 “.250 finer .954 82 28> * 250 £80 62 25.2
...... .oEOnmm
Eonoqum
EEEEEUHN
.mswcuzzou.
.2 >55...
53.... umemE .5... 5.585 5.6.
.6 E25? 25.855 335: 2 m. 53% MES...
.o 3.9 .23 .253 5:836 .35.... 3....
.3533... 55.8 .0 .32 a 23.5. .3555
v8.36 2: ES... 8 Bozo. 32.x. 5%.... .3.
3.2m .o BEE: 3.8m E ...on emu» QESE 2.x mum 55.x 62 Q. .55.“:
33.3.. 3.5.59: .852 .8 62 US. 85...... .8» 832.. mvtv can: 2 2n. .02 n: .352).
owSENE cosmos? .o 65». ow< 3:20:23. xum w:..xc:\w:..v_um.._. .osEuE 22.3.5; .o 0:52
aux: mam: ham: 3;: max: XE: mam: «mm: 5::

 

35.2.52 22.32: 59:5 .55 2...;

.2283: So» .0 5852: 52:6 05 ..o 5.55.85 5...» m: 023:. 8.85

. .529 .5 =33 938M332..— .8 “in: Egon—Sm save—.381 .onuah

gig—NEH: DQOEM—mDO: Egg—DU "mam: ZO—Humm

 

209

 

 

   

    

8&3. 3592 no 38.". no
EEG 9:3 as s 955 nu 2965 u
886m up 5.5 n
35:93 .50.“— 5395 3.5% no v05 u
3.550: E55 um ........... 550 H mm 3530 um
EU BEoN "v ..8 3:8 on 2 3:33 v5 xofi u 2 9352 uv
55 352 um x83 3 to 23 n N. 5:826 .29.: um
........... .550 H mm .35 96:23 "N 3:. ..onEoE DEE» baa ham 2 quU n _ _ .cuE>o_aEu a: 83... "N
has—.8 “0505‘ u 2 E0 9555 u. 2.3— wcioow u o. EuanEEu ..2 533 n—
:o. . . o c . h m:

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

250 ..8> 350 250 ..3> ago 62 e aEaZ
no. @8630

QEEE 2: £3. 2%.: $5.. mwfi 95E ..unEoE o>oE ..onEvE Eczema:

A.,Eoc BanE 2: EU 2: En 20:3 2: En EPA .553 E3:

quo ..onEoE 2: can? .votsooo azEE 05 69:38 3:3 8:38 ..unEoE mE.
En 20:3 .Uotsooo :EnEozfiwE. o. .5225 2: 5:898 .:.auu\co_§m_E ..o 5.20.253. .cZ BLED—2
cosfiwgé : -E t he cemaox So: -30: 2: L8 .533. n: 52:52 20530: .8 oEmZ
Eat
my «it 5:25;. £3 S 3253 3: an
Eozvmzo: OH :o_:UU< 20:03.01 EOC cocombnzm 0:3 {MANENEKQ hez QN ES» 535»: ,§~=U_

cum: _ mum: _ 5m: 8:: a 8m: A 5:: _

 

 

35.352 22.88: 5 8326 on: «zap

\ .33 333:8; a»: :3: .3 E2 3.3 in: :3: .232 3 5.53%
@353 a a we ..iEuEke tum x3223 S 3933a: 2m: 3% EA uQESKS x353: 338.33 aka 23:32? 3 Vufixmh h. 5.52me .Ewafiwfi 39333 E: 2 338535
B 25:33 2: 953 S 3522.: En ~ $9325. §l§< Esta 3. ES .993». «.OEPxESC :5me $853 «Etnafimfi 3334:»: S ammSEu 3mm: 5.53. 3.:

...0!» .5 :3... 93325.32:— .8 .52. 29.3.5: 3:91:38“ 335.

filmy—magma: GAOIHmDOw— 7: mflUZcfl—U "04:.— ZO—Homm

210

 

 

EfClrv avian—"mm

B§ZHNM

 

..ufiOflnm

Econ—oar...

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Egon: .oeoauqaonom 5.385%” 32-3.55..qu Setsfifia .8..3 €52..qu .983 25:3 .33 95.8?“ 23...... 83......» ..omuv
3. 3.2... was; ...EEum 55...... 8.8.qu ...aEOnN =5. 8 25...... ”555.".
$32.3.” 35-5.2.8...HR Egfizufi 822nm =5<u- genus
gag—um. 8.30 589 no 85 «a gush—=89 :32
0.50 3x5 260 £50 £50 £50 0.50 oz 8 when 250 9.50 on... 23> es. 62 25.2
38d.
vats... £38.38
539 32.... 3.. ........
DOM—m ho Duo—#3 QHMEUSO
3.5... 82“. SE a 8 mm .w :2 a. E. ...sgsunn
. manuo .8- 558... has»... $.83 3358..
v.85... .2382. 28 £152. .3: 83.... 95...... o
...................... 355%.... 4. . .1828... .85... 232. iguana
550 "mm fumes 3.8» .235... mama .5 282.. 6...; own Benn
8.... 2:2. :23. Ho Fan 6782.. 9:538 inn-.8 no... m9... 5 coco... oEoSwnv
.358: .53. um ”£38... .22.»..8 ...zumszes .232. -3. $2....um
83 .83. 850.3 uv oumficau. muojuauno... on... 2.. .8388 .5208
9.95» nuance—33 0".ng
.838: Hashim Ex“ .nuficano... .25". ("A333 mats 05 @222 323
88 ..o>on_ flung. Ems? 296m .8538“. mm? was 059... .338 m3
888%.... .meEEooHN «"3887 05 ..8 be... -..ocum
no... 53...". 5.523 3:55 guess... 9.38 39. .5208
8:55 253:5 323.3,. 33 u $.35: 05.2.". 58v
8.26:8 was RES 9:. 2: 3.5 $09.82. 2: 3 .0530: mm? .8 0:... a
€2.88 58.. 62.8“ 85:3 $8 30...: mo ”589:? 05 53v .0 2.8.. 5:8..qu 58n—
mmuS: 9.2.3 couscou— ..o_.aE 58 o... 3:83 so» 50 E of.» 8:8 : ..o 3.50 Eu: 3 om< 5.80 ..o 9......
25:: 255:9... nEmENEmE. 55:29:. 8:: «9:. SE. 3:: 3m: 3:: 85.

583.52 9:.» he can— 2:3.

.320 .3 =33 3.1333305— ..o 18: 32.853— snovuonoom «munch

50:030.. .8 motzw 383. 05 00:7. Dogma 539 352333 5.88 mg.

imp—hm! E<NG 5.x! ZO—Homm

 

211

 

 

 

 

 

 

 

 

 

 

 

 

.02 a

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

”Boa—um 3.35—r.—

 

ocou . 2.8 £50 250 350 M2 62 0.50 M2 :2 was
«:83:
“>00 x: x0885
OOZ . 0:33): 5:550 DEE; go 039 3:3 x0883.
8:39:00 2225 03:; 3285 .80... .02 .3 09C. .955"— mmosz wwfiEsoE
R.N.~ wzfiwéggie 3:8. :0 5
95:8 3 .3 323334 .618 83.32%? St» 233 ES
Eotxfihnou wing 33:. humane 53.x: 20:33:
05
............ uDJHOHWm Us whmnu
Bram wacuzncoo H» 9.353
wficflfioo Ho cuquonm 33$ r 5% has
wczowasoo um miflv "Va 8.22 :55 8: Bo: do» :
cozmtoamaflkuv 02 um moo—tum 3% Eugemuug x: HuBuucS?
:Eouum mo> u. “new" St SENS
ween—um 238"; 29.820; £5 62 8 mg?
5303 $62850; an 09:35 333% 9538
2: 3 .2: Ana... acoEobo m:==..£
~>O\OOZ\8_§anU 8233 Cannon 220:3 «mom wctau
.ozaonEoSzo 5E=EE8 52.5 ESE: 85 22.8.6: 05 2 mfimcofin 98 .828 £85: .580
83038 ES 058 acousnEcoo :23 go: : =a 083 33:3 3 853mg: x8333 :32 mo uuuefim 35:3 En.
mug «Nam: fig cum—m: 39M! 39mm 5%

 

 

a? u: 5.9.“: 03.3.

212

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

                 

 

 

 

 

 

 

  

3 3 3 3
:32 0c #82 :32 3.32 :32 £50 £80 350 E3?
.550"? _.‘
wEu—uE 235cm EoE- mac‘s: 8.500
530% aflu MEBEM 0352 998 20c 5:280 mEEaF bane
850"? “580352 mica; 08.38. “mom 36:83. scan. 858 9:555
EEoEmomnn
Sin 3302: 8: Mn 35333.5
moxfl 829:: "N ..o 3&5
80295 H3323“ _,
E37. boézofla 3qu «Ea—Emum
9:an eunuch—um
uofiEoonm was Eu: 0:02"—
E 8:95 9:283 can wEuRE mEeEfl .
coon :omfloou E 8382. E 3533 mEEw: .
838% 05 we “Egg—9536“ ES» .8 aha—nut? ,
«fiaaozmau 8&3 is: «939 a tug: $28 :25: do» 2 05 mm: 05 88 so» ow >5: 538:2 35.8"— ,
bag cmflé many—m 3.2 nan—mm «mg “mg ennui nag nag

 

 

  

3% .2. as 3%.. :35 a: 958 3 3.8... 2 “a. 2.. «5.3—u .338: 2: 3 2.53350 «seam "at 3 55m: 2...:

213

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3% .2. as 3%.. :35 a: 2:8 .o 32.2.. 2 .3 a... “5.3.. 3.889 2.. 3 333550 «seam no? no £55 2.3.

 

 

oz
350 £50 non-5.2 330 ad u=32 250 «woo «too 0:32 maul».
SEOHnm
« .......... E96
Q85.»
.. EoEQMaEE 33V a :3 o:
SEOumm no.6 E t (353: -zazku
Baa—LE8 5389 m ua_:o_:o:nm “is $5283?
Rue—sotwn ...0:— o>tom 339.31% Eofiommch S: .8 L353:
can 0.. 55mm: .5 5353: oooanohnm 2: \o a: when
3:5 “o 2:55 . :ofiOumm AuSQVS .coEowafiE oZum wo>n_ V5» 8.3: 522.":
E oREum oZHN 2 7.3 cozmtomno ..8 ~23: 62W— moza m: :8 do» =
3322"” 8>u~ 05 mag“. 3 wfifiaflnm S .8 sum—£58
Eagouu— 25.3 no wu§:n_ 953 3: smutoucoE E fienEoE $233.2 32x8 :25qu
8:55 w 838% £323: wctau 3.35 39:3 02 8 mar 85
52?. 5388 2: Eu mac so $3 053: MctoEuE gnaw U033 3328mm
552; 9 >9. mmcuuoE 08an 22% ~58. Ev Ex» 5 339:: 333% 05 .8? 05
mm? :53 Bacon :2”:on “oh—:58, 2%; .5 9:58. now—30% 5:830 05 .53 mEov 9.8»
3% t 05 83 has. 3o: 2: 33 302 EU 0:3 3:3 .3» t 05 33 BE 0:858 20 .8 5252
e
FE cam! mg 351! mg «g Hay—m— ; «an: unam—

«Bo

 

 

214

 

.................. .860»? :35: 233m”... :33. .3255"... :3:an 96% . _ ..a :Ehno.
win. ..ooMIa 9.32. socmum 5....u:_...nh minimum. F:u&w0nm Nisan—J. .88.... Ethan.“ ...o... 13...... Bafimun A353 .55.. an... :5. ...0".

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.5239:
3.2.3.. 230 MS. £80 2.3.5— u: 62 2.50 0.50
...0» ...fiofio 5:0...
.2. 2...; 2.303.... .2. 2...; 8:8...qu
.3

25:3. 0922...

u... was 3...: .8» m.

ocum

8%".

3:33.55”.
c:....:. :o 2...... EgoE ..an
23E. ...3 one... :33 3:8 8%".

...:o..........:oo 15:5. 6:2... .5 9.3.0.. ..onEoE .995: EEK

omfloé o... a...» .23 do» .. 228.5: .528. -Ssaufio nun. 2:25.2an
-:o: a .3 22.35.. .8 .583...
o:nN u... 8 Eu. 3:2: .2... 8:925
3.". 8 35.3... w. 853.83: 35:30 2.9.3:. 9.3 DES: :5.
...805:.E2 2.8:. .8» Eh..-:>.o to :. ..o>.o>:.
macs—cu... 3.8:. 3.2.0.60 w:.v:om :. uo>.o>:. .2222). .2. .55... .to ..o 2.... ..omaouo:
:. ..o>.o>:. 8:88.. o... 83 .6808: o... 83 omSu>< .o .0952 :33 .8» b u... 33
mg _ 39ml ma _ «ma Ea own—mm «Xi: wax:

 

on: 359: has}... no «5:932 was. 05 "5.5.. 1.9339 05 «a 303—35:00 ”35.3... :92 a? ..o 5.de— 93:...

215

 

 

 

Ewséaoznfi agééaflufi 32-5.buau8 32.5.koeemu8 22.5.9an 822nm using 225":

.aéCunn avian—"mm .8502flm Saigon—m . ( <e t...u.._1:ECI0o ?
mag

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.BnEoE
E5. 88
2: no @2323?»—
uonfinn a Q 2.32
250 a. 232 £60 335.23— 930 £50 £30 £30 280
band 05 Bio?
0.55302» Mafia—OB 30:8 3238
...Qntoomwv .5 wE>= ..onEoE DEE 9a. 9.2 mo :5 .522 Ena—

 

 

 

 

 

 

.....550 3:5 EuEmBéo: < 05 52. 8 088 9:23— uoazzso $3 3:?— uohz EcsEEoU ..

 

:n a 8: .o: um
2:32: EH8 use a 3 do» "N
233E EB team a 3 Eco Sn .8» "—

EunEoE 3383 on“ go 58c 05 9.36:8 $5»qu beams? m8. .692 mcgo=8 05 he ha 3.: Boson—.5: :5» ED

 

 

wen—Hm meg NE SE con—Mm ama awn—ME small ema

 

.uonEoE « .«0 Snow 05 9.38:8 an 3 Ho: .5 «83.88 non—awn? Eh: mac“ .8 team an mammogram can: on. coho—n80 20:83: Box a 323 8 83 2.53 03

.58.. €2.38. 1888A— 2: 9.32.8 228:2. 3 33%: a... .23 .a 83325 8.2352 a... «a 5.5:. ”zap

216

 

 

 

 

 

 

 

 

 

 

 

230 :2 250 230 0.50 280 250 250 250 9.30 8:2

 

 

................. MDEOHWW KW Q33»:
50 Egan». 33.3:
2985i.” 30050:.
. . . 208"” 8.65? 2.00835 .502 £05
...92808 08am300wu~ EoEEo>0m 2:002 80c 00002.8 E00530:
Rio—0E0 $0302 >288 05 Eob Soc 50:05 Eoc 0.8.:
525088 800 0023 83m .388 £88 088 08:08 .3005 Efim 3:2: 300:.

.oaoumn 2.88323... 3% a 832.5: 3:080 8350 8280“ 328 33280

 

 

 

 

 

 

 

 

:0 E .00 .00 um
2:0on 0:2 M00. 0 ma .8» um
2032: E3 v85 0 ms 3:0 :5 .8» n~

$0052: 003080
05 00 580 05 330:8 850on beams? $2 :2an 95:03 9.50:8 05 00 >5 8: 32.800: .50» 0.5

avg aha aha 95g awn—mm neg beg ova meg

 

 

00082: a no 580 05 9:30:00 3 “a 00: 00 85808 couquE EB“ M02 00 :05 mm 8303.0“ 09m: 05 0032080 205300: .50» t 305. 0. 8:— 2:03 03

.5300 0.03808 03.3009 05 «530:8 20:300.— ? eomoifio $2 3300 win—.53 he Ramon—chm guauumz Q? he 55.9%! 03:.

217

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

...BEoE « ‘8 58: 05 230:8 E: a 8: .5 853:: 533%: :3: :2 .8 :2: ma wofiofium 3am: 2: 3:2an 228:2. 5% t 30:: 2 8:— 283 03

 

.58.. {339: 8380: 2.. 2.22.... 2:33 3 :8: we. as” ea $3.38.. .3. ansfim snag: a.» .o :32 as:

a e a a a a
2.50 d 030 d 030 a a 030 a. 930 a. 3.50 e a 380 Q 030 “a 030 250

a

0 5o

:fiamaE :8: m:

slug $8395 :30 >082: 320:
8» :3 55:8 mum: mEEE: 9555 3:39: was 0:3 :5 E055
H.53um 23: Eon wfifig 55:3 352 56:23 80:8 .8 2:8 3.83:9»:
Eu 2 EoE 33:03: as: E028: 33:55 2 :9. So Ev. 9:22 So...
min— 3535 355 30:82 3383. sou—m .329 gum : 3:5 < wEEab 80
25m =a a: go: .o: um
3.2 05 9582: E8 m:o_ a ma .8» nu
@332 E 2:ng .52 team a mu 3:0 :5 .mo» "—

3383

:0» 9a: 9% 05 E =u
EoEmmomud 9 mEEd: 20352: :0 Eu: 8 5 2:3 055 5580:. wfizta 2: “o H.585: n: 05 23» 3.83 @382: 338:
:5» E 05 no 58: 05 336:8 83mg“: ~09:ng 82 2:“: “Eu auto—Bos— m:_>8=8 2: Mo .8: am: 20:88: 5% ED
95%: «and: 5a 255.— an: ”be than 3.9%: pg 39%:

 

218

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

..................... 8:5"? 8:5"? Beounn ..ofiOunm
3.8: .3865. no 252": 2:82.. .85... 8 be": .85.. 08 2 32.35:: no
atom 23::— nn game: :2: 2 5.3—um 2538: 5:55.58 :0 beum 382:: 05 2:32 83:5. um
383. 38:62:»: 02:33.56 a 8.83 0:557: 8: 23.3: 8 be": 85.6 8.220: n...
502nm 825m m::8::oonm : 30:: 9.3:: o. :msofi. 305. .0: moovum :28ng .8 89:02 an
6.6.2: 88.5.8 82 38.32": 8...... 2.525": .3. 58: :5 9 8.2.1.: 8.. E880 u:
53: :o Fug—2:50"— n:o:m 23.5"— m:_8:b:::.on_ 88:39.. "—
030 030 030 £30 0.50 930 030 :30 3.5 930 030
8:85 8:85
8:2 £2: :25.
waltz: 25:2 2.80m 5:2
\Wfigu ............... .WHQSHOIIIWW
:2 r53": 3:8 Smog: 00:83:: 88:8:on
o:u~ 932:0 Scum Scampi 3.3 8:838 8:":
moan: 9:02:08 zomgawfii 5:25:58 8:52: wEEBflo
:23» 8:08.? b3": $53.58 8.3 3:8 oZuN ..3::m:::E 083nm . . . .. 3:5"?
N. 8:: x: 208.38% ~33: 8:": 2:: 98.80:. 833 ::2o:Uum
05 E .m 8 _ :0 2:8 22:39.... ...80: 8:6 8. ..Q. 3.3.32: ...Eo:8:o: 38: 35:3: 803nm £035.":
mn=< :o : :0 $5888 82 mach: 2.8.3.. 8.3 3.3... 33 So» ::3 mzouuu 3:55am
3:: 3:: .59: E 5:538: 5:553 0:02": wn=<u~
8:28 :w:o..£ 8:82: ...mtoto mn:<\>_= ...38:8.6 mesa: .8882: 25.8.3
820 :5 w::8w 8:5 :32 t: 38:86 :9: 38:86 2.36:8 0:. 85%: 8::
: 305: 5 8053883 9.0.3:: 5:55:80 8: :0» 9:: 2:: 8:8: :8: 20:88: 3% E: :0: o: ...:8: £5 :_ 88:86
:0: on :2: 2:: 0:3 8:3 .33 .02 = 8:3 .8» .= :9: 9:: 28:2: 0:23:28: 8:? 5:8 0:: 2: 52>?
asp—Moi 8:555 :26: 3:55:80 totm 5:555 mn=<\>_:
:2 3mm! r away—E _ a: hm: _ a: _ may:— a: a: «a 3.5:.—

 

 

 

 

 

 

.32... .5: 2.. “.32. m. 32.8.8 8 :8: 2388.. 2: 2. :32: 38.5882 $.53: west: ”:2: 2.3.

...0...» .8 3.6: 2:33.232:— .8 :3: 3.58.8: 3:01:38: Human.

820:8:0: 32:: 885.83: mesa: 8:83:80 :0:on 2:.5
wmmzmm<>>< ma$>=a “m3:— 205.9%

219

IIIOOOIII

   

 

    
 
 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

....... 550“? 0:50 5:5:uo. :05 525;": 52.5 :::E 85an 50:5 :55 80:":
2:232": 2505:": 5:832. 22:55: 855:": :2: 232
0
2
2
.02 .02 .02 .02 oz 0:00
28:8 2:: ...558 58:08
::::EE00 2:: 03:20: 2:: 0>::_0:
:558 :5: 5 20:8:0: 5 20:8:0: 208:8
”0.5305 2:. E5: 2 2:55.: 50: E 5% :0 55:2: 2:: 802: 28:8
0:: :0 50 20: 0: 0: 0:: 25:8 : 0: 0:: :28 : 0. 3:25: 0:: :0: E 2:. 0: 80::
82 :0: 0>:: 3:: 82 :0: 02:: 82 :0: 0>:: 82 :0: 0>:: 82 :0: 0>:: 25:50: :0:
60:8 :::E 30: 2:: 5253 3:: 5253 8.3: 9:253 99:: 5253 2: 52:55 .02 5:802
:22 60:8 :: :55 30: :::E 30: 2::E 30: :::E 30: wEEbc :0_:>> n: 59:02 20:85: :0 252
ohm: Max: Fax: 91:: «:5: CE:

 

 

 

 

 

 

 

 

 

82.53 5.359.: 353254: c 5 2.2:: 53::

.520 5 :35 03:035—305. : ._0 :5: 20:85: 0:: 2:: 023.550 03:80: 05:3 5:505 2:23—9:— 3:0::050¢ 5:5,:

_ ANOOQSONV 2:5. 0: 50m 5:080 E0: 558 25:55: 25.50 2:: MES: 85:50E 20:85: E 8: 0E: 80:23:80 5:08 82.:

MED mam. "Wm—m ZO—Pomm

MALA WI AGRICULTURAL PRODUCTIVITY AND ADULT MORTALITY SURVEY
Michigan State University/APRU-Bunda College
CIMMYT/MOA Panel Study Follow Up

November 2001 - October 2002

AGRICULTURAL PLOT BOOK: MAPAMS —2

221

SECTION AP-A: HOUSEHOLD INDENTIFICATION

 

NB: Enumerator Notes: Please complete the appropriate shaded areas before the survey begins.

I. Allocate garden numbers so that number I is the nearest garden to the homestead

222

 

39>

 

582

III as

 

 

”Snag m5 \0 amazon 2: BER EFEM {new \c $.5ch WQD 2: 382 :3. amxfiamwmgicflfimmzw B: .m.
hmESQ not 3358 S: SSEEN

EREQESQ E :8 .333 S Rotations» 2.5m E53 .2963 fizz V2231 393 m3 wEEuQS 203 has ataxia a. 33.3.33 3.3% 2: \0 «SE 2&3 9.36.5qu m5 £83m<

h2m§3m<mz 92¢. ZO—H<U~EPZWB POAEZNQ<U ”m.n—< ZO—Bmm

223

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

F
an]
al.
Sm
E m
mom
we!
8|.
new
8n N
no.
E
8—
No—
8. _
£80 0.50 250 £80 ”30> 250 ouou 8:08
Embl
8.53..
8:
£2 m
s:sm~& ..............
6558 .550"?
a .3 n Sfimonn
....... EQLUQUM “Shay—Q
.ofionmn 3V8 .050"? BEosomno
Eomunv 39.32 F82 Scoamun
Eomvnm no.0 N @583 52:93:
Eonbun afienm .3. an...» “Ea—nu.” ow2=>nv
Soon. no.6 .25 3&2.» 5.5m to." Socbnm
nu no.0 sages ..o >20". uraum
98an 0.8..."— 3: mo 2.33:". 5.5.0
no.0 3:55 @957. 69:52 9.8.. 25 .5980 b8 8 .38....
0%: 23w 8.6 35> $86 a: .0 oz :8 5a: .5355 8535 5.50
m _ mm< N 5%.. Emu? 2mm... 0mm... wmm< 5mm... 0mm... mmm< 35.4. mm”: mm“? in?
Sigma Pox—n. 32kmo Zmomfiu

 

 

«coda—.53 ..oughunao 39.0 .33 «:Snuhsudo—z mac ”ZO~.H<UOA lam—59

 

 

224

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

woo
we.
80
Be 0
E
E
E:
8.». m
3|.
new
E.
.9. v
vow
0.60 3.50 250 250 9.3% 300 360 8.32:
395
3.9:.
3.
.32 w
5§mwua ..............
6.5.85 ..ofiOHmn
u .3 V 238$
....... NBKQQUH ............. “BHBQ
..ofioumn $.58 .050"? BEusomno
Eomuuv 33:2! Ego. acéum
=5me no.5 N guano 5.598:
Eombum aflofim 2: an...» 5.83an omo_:>n.v
coco". no.5 .2... V2398 :09...» BEN 225nm
"N no.5 $35 as ..o >25". at?"
9.5an 28...". 8: ..o 2.33:" _ 5.55
no.0 3.53 3.55 53:52 go» on». cacao ..o o. .38 ..o
3.32 2.8% no.5 boE> 38.0 SE .3 oZ =om Ems. 3.3.2.30 85.55 5.56
_ 5
8mm... 39?. «mm—n? mama? «mm—n2. «an? 8?? 2mm... w _ ma? En? an? 5%.. v _ mm<
32.50 How—...— 32.50 Zmam<0

 

«3.5» 53.3.5: 532.536 RED 33 insg: mhfi ..ZQ~.N§.~ ZMgO

 

225

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

sum.
E
E
.loun
%
”F
owl
Sm
P
me!
New
.on
280 0.60 300 250 33> 0.50 300 woboE
\Nouo
.3qu
3.
3...: m
guENMN& ..............
SEES .ofiOnmm
o .3 V caucus
....... EBKGEUH HOD—.95
.28"? $8 .550an 3825ch
Eowmuv EQSE Ego. Scoannm
Eomvnm no.5 N hogwuv .8893:
Eunnun aflofim a... .5...» Mos—sum om3=>uv
coon". no.5 .25 3535 coo...» oenm unocaum
uN no.5 122.. ..S no Egon“ oranm
manoanm once". 8.. no 9.33:"— 5.5.50
no.0 3.5—n conga 53:52 who» onb 5.50 ..o o.
852 98% no.0 but...» Enemy SE no oZ ...om :52 3:95:30 85.3.5
_
ommn< manna. 392. Ran... nmmn< «ma? cum—n... m n mm? min...
3?.an Pan 32.50 zmom<0

 

«3RD» 53.3.5: .33....»an E96 33 32.5.3.3»: MLO "Zoz v.03 ngmv

226

 

 

 

 

 

 

 

 

7.:

 

IIO

 

Oin—

 

ml

 

 

m
D
U

 

ml
OI
Ulm

 

ml<

 

 

82w oh- 88m o WEE"—
wotaom no 5&5; 02m mar—«om 2.23m wagon 02w maroon 2235
Eaaxoam 3:23:50 ..o 598‘— canm BE 62 8E caaxoam go 5954 Bozo"— E680

 

 

 

 

 

 

 

 

 

 

 

3&3 EEK 3: no 2:2:3335 3: You»; 33 main”: KS 33 hangs 2: $5.: 3.3 83% 3.3:. 23 g 53.5% $3» unnamEN
musk as \o out. max 8: =0 bamboo ~33 ES 203 33 5.3% «88%.N

.3832 52.5

33.33% 28:83.3»: 3; no: 33 5336

227

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

N I >
> I x

x I 3

a I >

> I 2

2 I .r

.r I m

m I m

a I o

o I a

m I 0

82m oh- EB"— oHIEoE

macsm .0 593 03 gram 259% gram 2% mango 25:3
Enos—8m o>u£=E=U .«o Swami 9828 85 oz 65 Emsxomm ho 5&5; Eugen— c250

 

 

233 5.8% 2: to 228:2335 3: You? VS» noin 2?» VS» 93%8 2: $3.: $5. 33m 3.3.: 3% V5» 533% «on.» 83.3de

umum 3:: \o out. RE. 8: co 389:8 ~33 BB “.83 V5» 533% £23m<
......................... u .838: Z cacao

3.5 $33.83 3.25.3.3»: 33‘ no: 33 53.30

228

meow

mouom

mouom

 

”SHOE

BEE ton—

 

2

a8< K 5923

 

330:.—

.BoE tom

 

2

no? 6 5.50

 

330:.—

oBoEtom

 

E

8?. ”m 52.5

3305

..oto now go

..oto 3m

hobo m2
.208 ton.

 

mouom

mo.— on

a:

whose

so? a. 5an

Ste 9% gm

hobo 3m

8.5 92
.205 ton.

 

a:

330:.

32 ”m 5an

Ste 9% so

hobo 3m

hobo mi
8883;

 

mouoa

a:

330:.—

32 “N 5630

hobo mam no

.55 3m

hobo m2
BEE tom

 

mohom

@3305

2

no? ; :oEmO

 

BEE tom

 

 

mohom

2

<mm< Jar—ICE

AUBEUV «0039—503 230503802 854‘ uO—Am mafia :0chamv

229

 

52> 283":

EofiOflnm 2000.73 H<nE00muo~ omo: .BWELHE <Z<mu§ Zoom—ml":
So_~2\a~ooo:um_ .835 2058—050 5553 Eo=ow=d>MH§

 

~an>_=D E0250":

9:5"?

 

_ m _ v73 EcouanE. cum—mum. Gap—6522
was". coho—EU 2.: o>amua

308% ESEEoumanm SEE?» Eagmno ag_aeaseoginm N53 2<83<um 2835"” ag_afifio ag_szi
coat/oi coacaxm ..8 830

 

 

 

 

 

 

 

 

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3E: 8a 8E3 8:5
250 ..o 62 ..o 52 ..o 62 8:5 ..o 62 no 52 :55» x5;— o—EE i=3— oZ oEaZ
53%
83? WE“ E; 2:60
:2226 a 23 9:58;. wEEah 2233 3qu8
“SE bozo; c a 858 858 cougmcoEuo v55 meetom 33o: u>:m_om
no» u>mw 222x“. .mzcoEmom »50 _E::=otm< 20E :ofifiéfi .650 820 2:03;
conunmcmwoo $8: 8238 wEEm: EBo=£ 2: v -_ motowoao wEBoze 2: 95:3 BoEsz 9 now
553 00 oz 00 »5 Eggs :0» “3m: :85; £5 3:5 »§E >5: xcnm NEE 2 Bo: ooan. :o» 92:. 22:5 252 $9830 32.50
to: 20.? m5: _ 30.2 T 20.2 _ So: :92 0 20.2 _ 8: _ no: no _
250 £50 .5on 3.50 62 2.52 250 oz 252
EEO um
oZnN oZnN E30
mo>u~ mu>n_ .to 3:03 25va
»:E£ Eso oocfimnm oZum
«anew SEQ Boooznnm mo>u_
.EoEufl m .585 293qu goon» :2: 8.830 8033—
2 wcofin 9 woo—on oEooon no» 29230 2: 0o 59:2: van uEm: E 5580 53 52:52
so» 00 no» 00 Bo So» 855 2: 2 62.3%; 355 2: 33 as? .52 I: 2: :o» v.53 n: can oEmZ 9.5:“qu E850
h0m< 00.5. m0m< v0.3. n0n~< _ N0m< ~0m< n0 _
86 8E oEmZ SE 8on :onEsz 8E

5832—0 .2.— 35388: «ow—ah

WozooN common 32.50 05 he 83.5% 05 mm $505 55.5 n3..— ofi 0o QEEooEoE one» 98 wEEah .Rfiotuoxo maize—38 .5308 £50
m0h<mth 90‘: “0&5. ZO—hUflm

230

00258020 0:2 9:5: 90:80.00 2: E :5 200800; .50» E 28003 0. on: 00.0 .50» 5 9.3.03 3:: :0» 3: £00 95:03 »=aE 30:

 

2o .83: 8: "2

2...... gonnm

$033. .82.: 82 n2

 
 

8:: aznom “2.82%

00000.: »0 005005 NE :98 »n ansoiufl

:28 n B p: "M:

8; 3 Eu".

8 2. 32. .8 a: 8: n:

80.53 82 u_ _
Iago. .080: 05.4..

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mo...— :55
0:2 0:00 0:00 :2 oz 02 oz 02 oz 02 P:— 0000 ..0 800
35.5.
E: 3::
0.5::
V2:
>0: <0?»
VNMLC}
E:
a. 00.
N0:: N9: 533.3
Saabum $50.: 0:00
hisocflm 00HN 3 505.5.» 00.3
u :8" mo»u . . :0 05000
SE :90 _ _ _ 8 “HA 8 ~0th mm .0" A muwv WE: E mmflflm 000m:
:0 000% 000»: 20300 000% 3:; _ n. _ n— _ 2 _ 2 0039:: 8200.00 0
EsoEm 00:59 :30 BE 2500:“ 20000 0008000 2: E ea: 0.000 a: 28» m._ 9000 20000.00 9 8:“: : 050E 203 $5
:38. $0» : :0» 20 080.: 00 oz 000 3200 30050:. »:E& »:mE 30; .80.: :0 0005:: E: woo. 020 20020 :0»
.5300 :30 08E ..00 J 00:: 5%: 5003 N280". .80.: 30: 000..— E: 005$
29: _ 30.: 205‘ 30.3. 0052 w0m< A be: _ 00a: _ m0m< v0.2 M0“: «92 ~0m<
con—0:2 00.0

 

£33000 «2.— 35000003— 8900.

30:08 0803 .00:00 £5 mats: 95:20 000 .20 05 @5020 2:? 080.500 »0 :8: 80:00:. 25000.00 852088 00:08 m2:
20:52.05: 07?..— .».0‘5 “0:: 203.0”.5

231

...mcuswa mats 335EEoo 05 E 93 20:88; So» 5 £833 2 26 SE 5% 5 wife? 3:: .5» Ba «56 95.83 ESE Bo:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

thHWW avian—=00 fismfla §<Hw flaw BEEOMHF yoga—Q GBEU heuugflo hung—QIXO NW
Ehno 3:80 Sequin—:2 coomuw ..thm Bozumuv 9:396 oE—uv 33338.5 um
.85 QEEHM oozun 283831303 852.. asfiaambafionh mafia Scum 3% 2.. 223 9:356 u.
g m U00 8:302 5.55
:2 250 0.50 :2 oz oz oz oz oz oz E:
houomhum ocu
zoxcovnu mohfl IA v IA , v 5:523
2383 3%,me moi—mum WW2 mow—E £5 E wwEEaE
...wEEnE he 95% not: 3295 229:8
9:; Ext :0 95% $99 biog £96 EsoEm Bacon :25qu of E :3 x09 35 £8» E 30% Bacon “0 9 8:2 :
EsoEm 38,—. 52:3 wmoz : 2E so» ED :22. go oz 95 323 EonEoE 3E5 acmE Bo: .82: “0 59:3: Eu w:o_
5 .6; £95 RE: 823 3:: 5 5 Sons ESE ESP >5:
@—
cwmmt. _ 3mm< _ ma: 2mm< 3mm< mfimm< 4 Ema: jmfima< — N§m< 2mm< Ema:
350 280 :2 32 MM 250 250 250 «5.5—m ..c Sun
cZnN 5:95va $333 when
mo>n M consign ocuN :wobnm thfimwwka 5:33
BEEfi my?“ $81 3:33 am: wweu 33:3:
253E mzsuc ABE: cosmtonmcg 8205a wSEuE «v.2
522 :2: :2; so 2235:“ so 98% E now: 83:83 ~.=o_:m_:com ..o Bum was
Sfizcfl Emma 83 mm? EsoE< anE< ©3530 coEuE Evfibi Mo 85cm :0» Eu @052: 22a £5 ENE
raga .258 ED 59: Bow @533 555 so» En can?
am: mm: 5.2 g cm: _ mm: _ 3:2 _ 8.2 am: a:
" 53% “max— 8E mu: 5 ‘35:.— nouu ESE—.2 8..—
...8225 SE 3:01:38: 33.;

30:08 53% Eats”. 25 he 6:... 05 95qu 0:23 5580 3 com: 82.8.5 EEO—Shun 853228 .8303 3F:
OZ~HZ<AA "Hg: ZO—HUMm

232

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

£50 030 :2 oz 250 0.30 oz oz oz E:
....... 559mm Env
Eczouaoaum
329:5 230nm :23..qu
8295a 5.8”— oZnN mo>n_ u . u . £5
$3» .3. 83 9.0% 8:: 2 925:..."— flvmofiEum 5 A 832 w _v 832 E 8302:
9:2: 35 E 235:8 3:95 283 5:830 2: E :3 x08 :2: £3» oaoug 2.35306 :33
@338 he 350E 9:306 :38 38,—. 1.0 oz 2 03% can 323 EBEuE szE has. Bo; 635 MO .3 59:5: 229:8 9 3.8
2: mm? :23 ,8» : own no» BO 835 3:: 2a:— 5023 was“. ESP : Eu mac. Bo:
:97: a??? wan: _ Sun—4x 3mm< _ was: _ Eur—t. H man: NSE< Sun:
ago :2 v.2 Mu— 030 250 250 250 £50 38
$269.3 x3528 \ 2&3 Petazmmu :93
9.353 a .3 333sz 8V3 N35»: .5323 E‘Bwfifi Emu:
:o \me99 3.2 numb tux nctaémac‘ an: .33 33 EEAA
“3:98 USE—gum BPSEE mi :53 gufiztwk 29:39:53 55. ark: ocnm
ocn wEQEv :cobnm .35. 3.: $555 .32 h .333 325 Rin—
moxv; .33 rigou— ..o~=:.:£ 5 Cha 3: wow: :0» wEE: 233.5%?
$5: :o 2226; ho comma“: $5283 u 9:396 2.8: Eu c052: gwuzwtmk 3.2 203 $5
SEEM“ wEmva no 25% 2: 5 com: 3283 $852: Ewan BEEQ mEmmEn cocwozngm $536 2 bfizze
3mg 2: mm? EsoE< EzoE< 5:0 EvEAmm no 850w .33 “o 2:; 23.55.. Ewan bags :33 bags
bfiztom £305 imam 35$ :0» En :23 so» So
E: a a: _ E: A E: _ eh: _ 3.2 _ a: 2: ~92 E:
33:52 BE

£89.25 35 "EouEx—uém Sana“.

.8203 538 2250 $5 he 83 05 9.38.6 :33 2...? 52:80 3 now: 80:85 0.50:2? 85:37.0 8:08 2F:
ZO—HkUEA—n: mug—aha”...— UZ—mmmmn a<m<m ”mg—4x 203.85

233

 

 

....... 850"?
«333“me
2522"»
$70

mos—mun

BEE“. Bo=omuv
522,—. 0335!”
002nm

nooofl

 

 

........... 350"?
983;":

«Baum—

oé- _~-m~um_

awed":

£=oEE< .8 223—=qu
Z<Uuo

8- _~-m~vwv+vEZuw

a 2.51.

95:22 5801.0

85:22 “momEoUun
2:52: 35.23
0.532 mum

Q ncsonfionvuu

mic” Sam "—
33%

 

............... .550"?
9690",».
EoEoanuv
8:522 gunman
wEoo: EEHN
magma—V85 "—

 

 

naocxcstofio H mm
95865—9 2:3 c928 EwB: 85. um um
mcamen

no. 5:» 3x2: :5 cozacfitow 5% u ¢
538293 .33 23 8 cozacficow 5cm um
B~=E8

@386qu 55 ESE =5 wanna—q 3 Ha
5:82am“ Emma 83 mm magma Hau—

 

 

 

ZO—h<0~:—.E< flmN—A—Hmm—h UZHmmm—ZQ J<m<m "rad—4‘ ZO—HUm—m “ ..8 $30 .3395. 9.3553850

234

wwEvooB war—6 bEsEEoo 05 E 23 228:0: 5% E 280:3 o. 26 BE 5% E watt?» 3:: no» Eu 93v mat—83 >52: Bo:

 

 

 

auscunm

moEBEomuo
Ea: E $683353 803nm
9:8: 95: + MEEESU eBSuQOnv
Eco 9592.5 :Bfiu xonm
mom—Ur @3383 :5 9:8: Hanna
8»? 9.325% wEoo: unann—

 

:ho£O nun

8893 Hocuv
3.83 o begun

.- 3

:33 o. “A

wcscsmwcnhhcscma Sam 33 m6"—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

wEtmmk
EM.
wEUBB
EN
35533
_:.N
:2 £80 £50 :2 oz oz oz 02 oz oz 95 250 250 9:335
5.857 ‘ 53:75
IN NS:
E3322
3:8qu ocuN REE;
228$ 3»1 n #nA n _ v n _uA n _ v socfiumo 33
8E tic 29cm 29:“. 222 222 2.: E as: 3875 EEC
:o 25% $99 $538 25% 3:; Congas 2: E can :02 :2: 3295 no £5 :53 MES: 3:3: as:
.5528 £255 fifiu 2E EzoEa Bacon ES» n. Bonn new .523 BEES: 295g ,8 229:8 2 mafia?»
=28. emu» : 30% ED 30% 3 oz EEE >52: Be; .035 .5 528:: 8:2 : 2c .6 com: tea 2:
530a cth 8:: hon J 8:: 39: .533 brim Eek was 3a: «550:. c8? :0» Ev 555
2.9.2 _ N5: _ :92 20.2 _ 3.2 3.2 _ 5.2 _ 5: _ mo: 5: 8: No.2 5.2
uonEaz SE
...3330 35 3:06:38m 399—.

3953 :owaum Eubao £5 waist SE 2: 9.683 0:23 8:22? 3 com: 8385 BEES? 3:33.38 5:08 25.:
625mm? 5;: 20:35

235

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2000 0.50 :2 02 oz 02 02 0Z 0Z R:
BEOHn n
:2..." oznm
:_O\:0_a:oonm 00>".

:BCHN 5:230
:28": .0:— Homfiww n_uA 812:0". n_v 022:0"— n:uA 0.222 n _v 3.02 confirm” 2:532:
:22: m:_::_00m :_ 5:25.: 30% 20:: :0:Eo:o 05 :_ tn: x02 35 2:00: :2 229:8
:0 822: 2: M36006 :2 332:: 120:. 2:00: :0 02 2:0» 2 95:: E: 320: 20:82: szE 2:2: 30: .0005 :0 .20 ..unEs: 2 8.2 :
0:3 8:3 .00» t 00: :0» 20 5:3 3:: 5 :_ 50:3 :23 120.: E: w:0_ 30:
32.: 3.34: Elm< _ C=m< E=m< _ 2:5: _ Ema: _ m_=._< N~=m< SEA:
3000 0:: :2 «v— 0.50 0:00 300 £50 300 2.00 0:8

9030...“:

9:25": H.330: Ectazmtc <03 Tuna
o:n~ :0 025 anmwtmx \ 223 Staging 3‘8»ka :_:wum REAA
0min— hfiEE: cgounm 3:23: 3:00 «5qu 2.: :3: 33 0_w:_mu_ 0:13 o:n~
92:: :0 @386 :UEQHN >928 Pen: «3:: uwfl 22321:: 5:321:25“? : 3:1

032:3: :2 :0 :08": $003 $2.50: 32 b 223 3:00 20:82:“?
BEE“: 50:32: $8205: u BEES: :3:th :00: 32:: 22323:: o_m:_m t2: :5
mEmmac :0 20:0 05 E :8: 03:95: .605qu Miami :2 mEmmBu :0» E: “0050:: $5.23.: PNQ : :0 8:3: 2 9:38:
:2 2.: an? EsoE< E=OE< 3:0 E023: :0 850m :2 :0 09C. :0:8:::a 30:0 :2 3::m :_:m : 0:3 :2 3:?
55:5"— :EBQ g... 52:5: :23 :0: E: :23 8» E :0» ED

5:: _ 2:: n 0:: _ 3:2 _ EA? _ 0:2 _ 3.2 3%? 0:: 3.2 5.2]

£3532 .0.“—

...8a..o:0 .2.— 3532—8: “awn-uh

@958 :82; E250 :5 he .2: 05 3:096:02 2E3 558:0 :3 com: 80:08: 220:9»: 3:23:60 :0380 :FE

ZO~H<UEmm< “MN—Emma ©2535:— A—OH. "EA—d: ZO—hUm—m

236

 

 

....... 82:0"3
:0_§00mm< 2083:":
:Fuo

mOZmHm

BEE: 323:":
8:8,: 235nm
OOzuu

:0oou_

 

 

.. . guacamm
325 83x5":
0-0-8-83:

“20$

...... eafizduo

s- 8-800122%
:09 um. 02:":
983nm

25 "N

222% n_

 

....... BEOunm
o>080un
60832:":
8:805; floumum
38; 055nm
3:80:08: 3
ago 00522
aonwmsmg BEES:

 

................................ :Bocxcztofio H mm
gnaw—:38: 3% 383 m 5:83:83 :82 u v

9.95.: 5:: 8.83 Wmhczommfl 93 £32. 85. c8325 um
9:352:20: 5:: 383 Yuanwmo: 85. BBXS< um
gag—:80: 5:0 93: E-w\:0:a£.Eom :0 0.83 _ 55:3".

 

 

 

ZO~B<U-EQ< “Had—91E Gamay—G m9: £d< ZO—Humw ” .8: 8600 ..oBm:< chad—8:830

237

wbEsEEoo 2: E US“ 233:0: .50» E 280:3 S 26 83 £5 :0 :53 65:8 @003 USN son 32:5 3:: :0» Eu 93¢ macro; PS: Bo:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

250 250 :2 c2 :2 :2 oz cZ oz in
BECHmm Ekuv oZHN
£OEozmcoQum mo>u_ 5:330
08 95a Ex: um Tao» ‘ I . . I . £5 E .320:va
www.coha swunwfl «ms 53 2.: 96% w TA 33:5". m _v 3:25". m TA 3732 w _v 882 339:: .w 2555;
255 E 82238 EsoEm 8:: 5:830 2: E Ea gas :2: 2%» 038a 2: 3ng
chscum mo @0505 .5 8229.2. 38% 0303 .8 oz 2 30% Ea 3205 309:2: 3E9. :9: Bo; .32: ho ho 53E? 2 8:: z
05 33 3:3 .3» t 8: so» 20 “on“; 8:: 59: .5083 g Eek Eu was >8:
5.? 8:2 3:2 _ Sr: 5.2 _ Sr: _ 25¢ _ 3:2 2:2 Sr?
250 :2 :2 =3— ovoo 250 250 350 £50 £50 2.8
Emma
uEu—Euv Fetufizmmu :uam “SEA A
0203.5: 23:8an \ 33$ 534:8 ESQ?» oEuzmoanN SQ: ocnm
ocuN .o :BSHN 8V8 FEE: 2: €34. a9: oEofibsfl Rin—
muxr; 02253 2803 76mg mm: H335 mwfi rotgzzmv‘ se.:cummtu 55.
$8: so we .68: 33525 $2 5 .323 8&8 $2053 803 25
032E; comma“: @8223 u mchsaom @2053 go 02038: $3 332.: :SBQQSS .o :0 32035;
30:55 so 30% 2: 5 com: 8283 .o .852: 02055; .5 02052. :0» Eu cofiuE 3.:wa 8,: ”.2955; 8 “2053
2: ma? EsoE< EsoE< 3:0 .cofixmm .6 850m no 09C. cocaaamm 13:55 2: ban“ a was ban
8:8: < “Eosmfimgoss: :33 3% Eu 5:3 we» .: 8» 2a
3.2 _ Sr: _ a: fl a: _ F2 _ 2.2 _ 2.2 3.2 9.2 3.? FE
flux—=52 .3.—

..SESQO .2.— 35338“ .095.
3253 c038 E250 WE. he BE 2: so 883 93 “won @5828 233 Sago 3 new: $5:an £Eo=9w~ $5538 5:03 3;:
AOy—HZOU Gum—3 a Emu—.— £i< 2050mm

238

 

 

....... £050"?
conflmmoflxx Euchamuh
mFuc

mOEmum

6:5"— Bo=omnv
$va 03> Eum
OOZHN

accou—

mouo . Co m 230m

 

£050"?
vagina
o._=80<uw

GD>OWHB

x2035!»

Econ—van

avg—298 $3003.
Buofiufifium
Eanauum
Ecocqifioctomhvu _

 

 

bfiOnnm

£935.,” ccazuc
>15um

Ease—m 223
608083 98: um
EooE mcaooam um
McEwan x332! "—

 

 

..................... .550 "mm
managing: .8? 8.33 m hocéwczomms Sc< u n

mecca—n. 5cm 8.83 w.m\w:=8§ Ea Ewmoz 85. coogom 3
wanes—am; Baa 383 $9.38; 85. 2833‘ um
manna—6‘2“: hot“ 928 3 -wEocmEEBm no v.83 _ 555/ um
maven—a ”.4. "—

 

 

 

AOKHZOU OHM? a Emma— uni< ZO—Pumm " ..8 830 .8355. 23583850

239

SECOND VISIT SECTION

240

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.550"? tone—Edna mEqu ..otu “Egaafim mm» mm 22.3"“ 580 o n “.2351"—
! «So 3. .oz .oz 52 .oz 2.8 2.5 wolol nonm— Idalm 2.8
£50 280
23% 38 2.23: buts» no.5 .2 =3 :1—
3»; Son .8 Eu; wEmmE. Boo 35E 9:233 38 Ammmmon— 39.899
.2.— mcoaom new we 02 ‘8 oz “0 02 we oz bog; can 35 §afiEooU $0 .3822 SE
22¢ «52 :E< 33¢ .52 a: 52 2.2 _ 3.2 .52 a up? 52

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Oahu—00mm Vggw EU homo DAE>

 

 

 

 

 

35302 SE 0522 («o 02 .v

222056 a; .m

.chuum SO coho 23> fie oEaZ .N

....... \.......\.....hU>H=mu=UQ0hobo “HMO .w

 

 

«8:353 mot—5 =8 no.5

3988 838 35:5 2: 5 0%: 5 385mm 22> 85238 :88“ “2.:

Egan“— .HDU momanZH—zgwaflz GAE a: "Tn—4‘ ZOEHm

241

 

 

 

 

 

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

02058.”— 63:00
“5635: 803:8;
m§£éE<
am.
hufizton— 25 BE
mfimmob Nflom EoEowaSw—z
wztaan 580 3.55... :om
w:. p: Em
a
3.80 uo>=o> countomcoo
mama 232°: =om
g
3
250 £50 £50 £80 230 350 58> 280
gal :5 EE H9:22 5.;
.oEOunm
wage—.663 $=o>oin§
2339: 82850": FED
Sew. DEE 8385 muggy—HS $3938.”
Eu; 832053 Ra
3:9: 2:32 2:3 5:02 mEo;uw um 83% qu
Egon 83.8 «EEZHN. 3% 305.305 ocnm \umum :0: 2: .8 \nwsncuizu
2383 2 033388 mac—no 05 55 coflomufi no» u_ m: was: 33.3: SRQVQD 32 met
$508930 oéum uwwtonm hen—Sum now—Sm :oz bStS .38
“SEE 2: no .8388 3:323 8223 203 £5 5:3 3o: abut”; .935: $33
9:23» Hozwwmum 9 >258 oz "a co :2: 85m 4:05:32 05 Scam F82 when burn;
9.5.532 .oNErénm 339m 30» >55 : 0209303 no» Eu Eon? 89¢ net x223
E328 E9496"; oocfimmmmm Econ—"fl 9539.3»on 2: we was >mo€=nao<§ot§ .
inc—occogbotg £5 Sosa 8.: so» coon 3.50m 05 35 main . .. ., mum—U
BR: “finan_ H88 92 05 8d Hun—>9 93>» 62 .3 :0» 9m: 33 :53 53 so» Ev can? a. . . awe—0:58p.
QM“: _ 33¢ 35¢. 35¢ 37:. 3?: _ Sr: «in: Cr:
£3552 SE 22.8.3.— 2: E 5535 .8 32¢va SE 3:033an “awash.

H 803288 .538 Eons 35 SEE.“ .3 255383 320502 .«o 3:52 35:28.8 cocoa 2F:
Hzmzmwmmw< #8407503 "fix? ZO—Homm

242

 

 

....... BSOHnn

coucaxmnm—

aoaaomufi

aaofioomn:

boo: couo=<uo~
8x38 oEmuaomua
9223 um
Huggins

83:688.. “8203
BBo..m-SOun

22625: 359363... 30:83
hone... 033.5 nm

 

 

350 H mm
Beam Bazaar???

5 22%2

23m"?

$59":

coacaxmnm.

fiOhflOwoMHN —

omsocnoomu... _

Soc couos<u2

8x38 oEmvaomuo

Um<20< um

aoxuflzflh

835208: .BEUuo
339m¢5un

22.88: 5503.55 Bo=omu¢
BEE. 0335 um

mOOZHN

coca"—

g

 

 

HZHEmmmwm< >00AOZIUE "M.m< ZO—HDm—m " .8. 330 .5393. 9.5.533850

243

 

 

 

 

been?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

...........:......_ofiounn maotan—ooanN— 00020.80": 38:3. 008 mic—av, 80¢ 300 mate—coma”
moficsnno 0.030—05$ floxmanuh 836E» monocsmum moi—"v mmxum mama «acmnm $3 950". 9500ng.

Sm 0:. ,8 ED 08 0000 000 z 0.08 :

55$

3::

0:0qu

3:5

31.5% 3...: a x: :2 e2 e2 A.z 2 oz oz m... 250 o: .53:
9333
.33 Ex.

.33 .auuwgh

fluA m _V EUA 5v 3.: 00.038

20E". 290m 222 202 «ES 323 .83 $5033

00 $5 .0000 ~20?3

0005000 80 35 0.3.30?

AtEcmaEEnnfiov 08E: 2.: 5 7580 gt 0.5

08% 0:23 00 000028.: 000% 00:; 0002000 05 E :8 x00. :2: 0030i: 0. 8.5 003: w

020.5 3 00m: 0258 3 00m: :0 000% EEm 20000 28» m. 9000 0:0 30.00 20080:. 20000 ..0 : Eu 0:850; 20...
30:2 00 Q5 08:0; 00 Eco E:0E< 220,—. 00 02 3:08 DEE 30; .08.: 00 50:5: w:0_ 00 080 a... .0350:
E“ ; comma“; 000 J 02:.— .0 E .5003 :Emm .309 30: 0050:. 00» 0:. 55$

20.? _ «3.2 _ 20,2 _ 20.2 :0“; 03.2 _ 5.2 3,2 _ 5.2 _ 3.2 _ 3.2 3: 3.2 3.2 3.2

 

 

”honEsz SE

3.0.0.300 «03 3:00:38: 390,—.

3953 0808 000:8 £5 war—=0 8.0 05 9.08%: 0:55 003.500 3 08: 8280.5 282.0%“ 85:0st 00:08 aft
UZEMEME "as: 205.35

244

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

is}: .02 .02 300 .050 0000. 0000. 90$.
7.000 00¢ 200: a
20800 00008 0 029.52 £6 8000.380
200 003 00000000 05 0. M00302”: 2.0000» 05 0083:0«0 0088—000. 003 00Eom00m
200000.. 00 000 00:02 00% 00 00000 000.0000 :0» 0000000 .003 00.0 05m 000 .0000Enm000 0E0: ..00
:0: 0003 .80 20 00:0 00:3 :0» 20 000.0 30: 20 0003 05 203 00003 000000 003 0000000 00 002000000 .003
32.3. 032.3. 3572 55.: 32.: SEA: :2“: 2204
0.000 .300 0:2 .02 .02 .02 .02 .02 02 0.3 0000 2000
tecnnm
030.50.803HN0
00000.30":
38000 0200qu
8:003“ o Eonum
woaxoium m
an ..
3mm“ 3.: sfiazum
80000mun 00 00: $00000
3003 00: 000000030 0000 00.2
mwxnm SHA 2v m _ "A m 0 v 0008000 20 0.000809:
mwan menum 800800 020.00"— 0202 0202 £5 E 200000 000 ..0 0 $0030:
”was wxoofl 30000 00000 0000 0008fl0 002900 00 8:3 000800000 2:
03.000000 000080 20000 05 0.. :00 0.000 005 0000» 2 9500 000 20000 00 0 20 0:08 M03385
0000005320 00 0:03 00 0300 0500. 00 02 30000 0002020 30000 00000 300 .0005 00 009000 m00_ 000 0203. 000 08:
5 000 08000000 030.0 000 J 00.0: S 00 00003 E000 :00... 30: 25 003 0050—2
«:2: _ SSE< 32.3. _ 020d. 3204‘ 02.”? 020< m20< 3%: m20< «20¢ :2:
“03502 020

£23000 .0..— 35000003— .0300.
5958 00008 000.000 25 M0000 020 05 @0320: 20.? 0008000 3 000: 800080 Mcuoxafi 000 3.0000000 80003800 00008 20.5
622924002 972 Uzumwmoomm "2.: 203.35

245

 

 

.............. 0050 H0
00000<uw

08:02.14.
00850000010020uo
00300w¢00um
20000003000000 Bozomuv
000000. 80>t0 um

OOZHN

0000"—

Hmzso 308.002

 

................................. 0050 "mm.
wcfiaxoguc

wEmmmmum

058.0%

050.20%

05:2?”

M03025."—

 

 

 

OZ—HHv—mcaz 92< UZ—mmflUOKL ”Sit. ZO—HUm—m “ 000 8.000 3384‘ chug—0030000

246

MALAWI AGRICULTURAL PRODUCTIVITY AND ADULT MORTALITY SURVEY
Michigan State University/APRU-Bunda College
CIMMYT/MOA Panel Study Follow Up

November 2001 — October 2002

AGRICULTURAL PLOT BOOK: MAPAMS -28 GARDEN MEASUREMENT SUPPLEMENTARY FORM

247

 

OIZ

 

ZIZ

 

2....—

 

Alv—

 

v—In

 

 

 

 

 

 

 

 

 

 

 

wctaom
00030—00m

 

820
0: 50:3
0300—2000

 

2:0
0: 50:3

 

m0t00m
0003000

 

0.0.. .0000
m000§m
.20

 

.02 020

 

0:0me
00030—03

 

2::
0: 50:3

 

w0000m
0003000

 

0 0.0.00.0
000005
000000

 

 

i003 EEK 05 :0 20050030005 05 0.00000 has 00305 003 350 00.00500 0S met... 0.23 .803 00030 05 020 000.3% {.000 05.30:.N

00000 0.30 K0 030. «5:00 :0 02000.30 ~33 050 .201 050 000.5% {0082.0
0000002 000000

000.033? 30.00.3000: 00.01 3: 050 000.08

 

 

 

 

 

 

 

I
{0

 

 

 

 

 

 

Maw—00m
2030.00m

 

002w
00 5ch.—
0300—0800

 

200
0: 50:3

 

wagon
0003.8"—

 

oh- EB".
002005
83

 

.02 8E

0508
E03005

 

 

020
0: 50:3

 

magnum
0003.50

 

SHEEL
002005
000000

 

 

20000 500% 0S :0 20050030005 00: 0.0000; 050 00505 0&3 0:0 0005200 05 050: $00 0003 00.38 05 0:0 20.0000 «000 0030000<N

0000 0.30 \0 00.: max 0:: =0 b80030 N003 0:0 .203 .020 200.00% {00.3%.N
......................... 03:52 000000

3.5 40053.5: 3030000003 000% 00: 050 0000.80

249

5:0 00w 0.0

 

 

 

 

 

 

 

 

 

 

 

Stu 3m
8.5 mi
00808 02085;
0200 0: 08< ”0 000000
8:0 000 00 Ste Q00 00
Sta 3m Stu 3m
8:0 m2 00:0 m2
0:208 .2085; 0032: 082:5:—
0200 0; 000:. K 50:00 0800 0; 08< ”0. 000000
8.5 000 :00 8:0 00w 00
850 3m :25 Km
8.00 mi 8:0 Q2
2205 08085:. 00208 020:. ton.
00.60 0: 02< no 000000 00:00 0: 002 ”N 02:00
8.5 Q00 :0 8:0 000 00
8:0 3m. Stu 3m
8:0 m2 :oto mi
00208 0225.5:— 00808 .805 to:—
0200 0; 002 "w 000.00 0200 0; 08.0. M. 000000
0:22: 02085:“
0800 0:

 

<mm< ..EFOP

25:8 .00:00—.83 0:082:30: 00.3 .2.— 000 50006

250

MALA WI AGRICULTURAL PRODUCTIVITY AND ADULT MORTALITY SURVEY
Michigan State University/APRU-Bunda College
CIMMYT/MOA Panel Study Follow Up

November 2001 - October 2002

ASSETS & LIVEIJHOODS BOOK: MAPAMS —3

 

251

 

2.3002 “£30m

 

 

300.3323

 

Ill
30.0.0:

 

 

030mm

 

00:

 

00......

 

9.00.8

 

.22..

 

£000...

 

 

s00...

 

 

mhmmwt. ZOH—bbflOy—L

 

 

 

 

 

 

 

0:... 0332...

 

 

00000: 00.00: 00.0—£5.33. 0.0.5.5

 

 

00000: 00.03— :00505— 2.50.5

 

00000: 0803— 50.05305— .50:

 

 

00000: 00.30% 00.5%.: .53—

 

 

33: 3.8.. 00.0.3... .550

 

 

 

 

 

whmmw< 02¢”..sz
0:2 0:2 0000 30> .02
..._00300th0 BEBE—um 00.30.0000
0030 8:0an .tOnm ..0
:00.Unm £000". 0000:3300
>000. .0000 :00 0. 0.03 .000 0000050000 00200500 ..0 000000.... 00:30
:0» .= 5.300 :0» 030.; ..0 00.0.» 03:00.5 ..0 0005...:— ._0 0:002 ..0 00.; 0000.02 .0000. ..0 09C.
o<4< 9.12 3.1:... m<.._< N<4< _<1.<

 

 

 

 

 

 

 

0.000< 0000000.:— 000 00.0000 E ..0 3 <1? 200,—.

000: 0.0.0.000: 3:00:008m .0025.

7.500... N. .00.. 05 .0>0 0.000000.— .3 0000000 0003000 0.00.003. 0:0 0.50.0030 .0000 0000:0500 02.000 05.5

mhfimma. .<.x—< ZO—hUM—m

252

 

 

 

 

 

 

 

 

 

5.26 080

 

050$.

 

0m E

 

08:0

 

$00009

 

 

00000

 

00:00

 

0.02m

 

00.5 00000....

 

0000 00000.00:

 

0.0000:

 

0==m

 

 

0300

 

.02

.02

.62

.02

 

:00 .00

SON
.00E0000w
0.

300 00030
000.02

0.00
00 00.0
.0 00:0

0M0.0><

50m

.00 000.0
0.00
$00.02

SON

000.;
00.0.0 .
000.02

58

.02

.02

.02

.02

002

.02

..fl.

 

000.0
00.0
500.02

Sen 000
000.0
0.00
.0050 Z

.80 so
00000 00003
025 8 0.0

00 003000.
500.02

SON

000 00000
02025
000.02

00000....“—
an
8E 05

Bow

.00 0050
000000.00
000.02

Son
.00 E
00030

.00E0Z

00.: 00000030

 

 

 

0:34

 

S<A<

 

0~<A<

 

mu<-<

 

v~<d<

 

317..

 

N~<A<

 

20.44

 

03:3.

 

«<17.

 

33¢

 

h<-<

 

 

0.3 0030033 E H0 5 <44 030,—.

253

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

32.5 :55 :3: "a 5.550"? . 5680 BEE no Egon—no
swims—c... "w Ba .6938 n ..... .25"? €9.3me
“9.qu EE. up :38. $9.82 N .8580u2 wisp—m ..uomuh
9:25: 82503 02 no anm u mEcozun— 230nm _ wEmmoa =Ouo
33053.. 3:23.28 .33 :oEfinEg tam um ...hoSOnnn 3:50 wEnEozm :63 um. wciooxoomui 3:035. Begum mEBom um
5:59:00 52: 8... .4 5:2: a 55 82: 8.. :6 950 uv 283 n 9603.. "2 SE; “.85 "v
3:3 975:. .6 x3..— nn SSE a 2 x83 0:0 :5 950 um ows=> n 9:55 "a. 8%. own? 23mm um
noun—.538 0:333 5 um 050: E9... .23. 5900 "N 323:2. um €595 u_ _ .538 Etc :8 3E: um
............... 850 H mm coast—5E0“ .53“— oEo: .u u. :5.“ wctonaflucfl .EEmE... HE 3233 .53. .330 u—
E o_ E V o c c u o 83 83%
£50 :2 PE Ex £50 £50 350 58> £80 62 a oEuZ
...NooN
5:33 ..onEoEum :Homnm .02 n: 82E»:
3V8 3Q can 538 82:82 20:33: Me 0:82
N. 58 52290 @558 333
EoEonEu 8:3 .tOuN 323 $35 v.8; 53:8
Ehfiéo E 05 MESU 95: cows“; Show Sufi 588338 £5 E TEE

mEfiEoEma EuEonEu 33.535 mFEtfl M. .9355 3203: EVE me Noon BLEoEem was
E 53:85 Eufléo uE: 32E «8.2a wESvn. «5:830 chance wEsow Sen 53200 c8352 3.23:8
:0» E0: 5w :0» mEquE EoEzoEEo Mo 25 223 Sam ~.uu>_o>E oEouE E5. to E 335E
on $5an Ev 0:52: .3333 ES Eufléo 2: ...ocov 83 So» cosmos :0» Eu 9385 E5 coon mm: :2: 5382: 20525;
:23 20:8 Bo: E Euqm 08:. 9 85330 E a 555 E 2055 05 E as? can? .to me 09C. 2: go .0: Q. can uEm: BEm
Sax—«V 3 md< any—t. wmdt. 33¢. wm< max—«V 35“ mmx~< det. _ mmd<
255E 5.5.. CO e ..c : .md< 2nd.

an»: Eon—8:95 "Eucaonuom «awash

nus—EOE 3 ”man 05 “26 228:0: .3 cowawco 8258 oEooE finance 3:33st 5:08 wit
mH—Eihz. UZ—Hém—ch @2007: "mid. ZO—Hbmm

254

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

62 .oz 33.5.— 3.5 0.30 x2 0.50 fiuagv 0.50 280 250 250
...2ufiOunn
Dun—Sum—
arran—
<2<mu2
mE<no_
u ....... BECNnm 2 5.:
cotmxuzfiom 233 cm 02
no .......... ..ofioumn oooanohua .4 2.25.5»wa
23:3 >282 30.59.8552 883 230.3%
Egan—um €828...— _ 293535 oZuN
x55 _ 33:3. eooauo mo>u_
.. . .2050"? 32058002 86:3" m Eotocnm
225.4 <2§uo 835"” @833. 88”
2820 2:85 .532»-§Oun 887.32 muumflfiozum .628qu
538 SE team 5&5 8.53 :28 wsfiaaoomuw I 58
:o» 02:92 Bo=omuv monocamWn not 25 23:20.23 .3900
Eu 28» xooao>=u~ 83¢. ”Mum“. .832 $5
.922 n 33 2832 39?. um as axonum 92.2 EuN $8 $620 cosonnoa
2.3 .5» 05 2c 3225 882 822 858% as «:8! 5.6". 8323 8 3382 3.2.3:?
Ba .23.: So 28» .6 agon— ES 038=Eu 62.6 .620 5 Emacs 28 .620
39. .28» .288 22 05 Evan—2 E020 ..om 59: $0382 33 5.20 05 33 83 59: 20 9:82
085 .5 30$ 2 35>? 05 3 Bo: wo2=om 520 no 33> E20 20 banana 35$ .8 0%... nob 8:3 3.: 3:3 .8» b :0» 2Q
50‘: g ~01 Sow: QUA< won—«V 51:. 90‘: p m04< 34¢ no.3. «04¢ no.3

 

 

 

 

 

68: 20:88: ”Eogoamum Hows...

8.288.. .8 52% :35 8:23.38 5.88 35
28:5 “ox: 20525

255

 

..ofiouo

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

onoflm
23529083
8.3888!“
29333 ~3on
so”?— 355 an m
manta
£5
:82 o.
8.8 : moon
2... mac. Bo: N
3:68
“830: 2:
Ex 9 85:6
05 mm 22>» _
£89 £89 289 0.59 oceQ 0.82 259 259 egg 259
2.5 E5 3:: 2.5 .25 .25 «ED 3.5 E5 3.5
“omen 3&3
98¢ “8:9: 82an 350 2:80 02% 80:8 32.8 .oz
35,—. “8302 <2<H Cam 0< .338: 0:50 9.62% mam Svcouom >885 880 8322 .8325 5.5
3944‘ 8.3. «a: how: 3‘: max: X54 ma; «nu: 23¢.

 

 

 

 

 

8... E8» 3 33a 32:8 s 32. 22.88.. 2: 3 2:21 352385 «56.53:..— .5 892252555 A. s .0 A12 03.;

2o 93?» 3 025a 2.2—2:8 .8 was: 22.8.33 3:06:38: “out;

72052.5; 3 23023205 Ea 803.2 §==EE8 9 $88 mesa—£38 :ouoom 3F:
EDSDKHm<§ G7? 30:53; EZDESO0 "9‘30 2050mm

256

MALAWI AGRICULTURAL PRODUCTIVITY AND ADULT MORTALITY SURVEY
Michigan State University/APRU-Bunda College
CIMMYT/MOA Panel Study Follow Up
November 2001 - October 2002

GPS ENTRY BOOK: MAPAMS -4

(To be completed by supervisor/researcher only)

257

SECTION AP-A: HOUSEHOLD INDENTIFICA TION

 

258

 

 

 

 

no

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

we

no

3

8

No

.0

.2553. 2:52.: _ 625.35 6.5.5:. 2.3.254 6.55.3

in;
«Emmmm m3 wSEVuzw 5:: SE 2: 3.x

33>. 32¢: :93 86.: xQZQ Kc 53sz 3: 5E\ :33 3 0.: 53:23:31 $33 «Smimfixe .2259 mi 59% 53:3 3 at LEE—E

$0..on .aEBaQ 5 8.55380 mA—O non—5.2 BE 80..qu 12.589 5 8351.590 me 56.36

 

 

QOOUm—m mmU How—.— HDQ A—Oy—U QAW~> "OZ—QMOUM—z DZOUm—w

 

 

 

 

259

Appendix A3

Appendix A3: A Dialogue on HIV/AIDS, Adult Death and Agriculture with Rural
Communities in Malawi: Highlights from Field Notes Collected during Focus Group
Discussions

1. Introduction

Focus Group Discussions (FGDs) are tools of participatory Rural Appraisal
(PRA) designed to gather information quickly for diagnosis before a program or project
design. Under the Malawi Agricultural Productivity and Adult Mortality Study
(MAPAMS), FDGs were designed to address a specific information gap that
complements the overall survey. Six focus group discussions (FGDs) were held, two in
each of Blantyre, Kasungu and Mzuzu Agricultural Development Divisions (ADD). The
motive was to broaden the number of respondents in order to give more people a voice
over and above the “original CIMMYT” households. It was also felt that in dealing with
a difficult subject like mortality and HIV/AIDS one couldn’t possibly anticipate
outcomes from an interview to structure enough questions to capture all issues. The
FGDs were designed to have that flexibility as well as to obtain consensus on diverging
issues captured in the formal surveys. This report provides results of the FGDs and is
structured as follows. Section 2, following the introduction provides the objectives and
methodology used in conducting the focus group discussions. Sections 3 — 5 present

results of meetings in each region. Section 6 summarizes emerging issues from the

260

FGDs, highlights key findings, and points out areas that need attention of government

and donor communities.

2. Objectives and Methodology of Focus Group Discussions

2.1 Objectives

The broad objective of the focus group discussion was to obtain a wider picture of
the current state of agricultural productivity, adult mortality and livelihoods in each
region from outside the households participating in structured questionnaire interviews.

Specific objectives were as follows:

i) to obtain consensus on diverging issues that emerged from the formal
questionnaires,
ii) to involve the local traditional leadership and get their opinions as key informants

and major players in the community,
ii) to involve different age groups and gender classes, and
iv) to provide a forum for other issues not captured in the formal questionnaire

survey.

2.2 Methodology

Two enumeration areas (EAS) in each ADD were purposefully selected based on
the characteristics of highest number of households affected by adult mortality. For each
of these EAs, all farmers in the interview survey were invited. Further 20 — 25 other non-

project farmers were randomly selected from the same villages to make up a total of 40

261

participants for each FGD. The village headmen from those villages were invited to
participate in their capacity as village headmen along with the Group Village Headman".
Extension personnel from the area were also invited to help with addressing concerns that
relate to their department. Also present were scribes or recorders who understood
Chichewa or Chitumbuka. Their job was to take minutes in English of all points raised in
English.

Discussions were held in two sessions, a large group session and a small group
session. The large group session was the main session that started right after the
introductions. This was made up of all the participants. The small group session split
participants along gender lines. A woman discussant was engaged to lead the women’s
small group discussions.

As pointed in the introduction, Focus Group Discussions belonged to the informal
methods of Participatory Rural Appraisal approaches. Participatory rural appraisal
methodologies are good at collecting qualitative data and information quickly but tend to
be limited because data cannot be analyzed quantitatively. The PRA approaches also
tend to be unrepresentative. For this study, care was taken to select areas of high adult
mortality per enumeration area and also closeness to national/regional boundary areas in
the case of the Central Region. Inferences from this study could be drawn, with caution,
over main maize producing rural farming areas of each region of Malawi. However,
Kamwendo Enumeration Area, Mchinji District may have some uniqueness because of its

closeness to the boarder town of Mchinji.

 

6 Group Village Headman (GVH) — is a village headmen that presides over other local village headmen and
gets cases referred to him.

262

2.3 Discussion Topics

The main focus group discussion covered several topics and the more sensitive
issues were reserved for small group discussions. A sample of the checklist of open-
ended questions used is attached in Appendix Al. An explanation of each topic and

justification of its inclusion in the focus group discussions is given below.

2.4 Large Group Topics

Wealth Ranking

Wealth ranking is a very standard topic in PRA because it is a good icebreaker.
The focus groups came up with four categories of wealthy ranking: the rich, the better
015‘, the poor and the very poor. The most common reaction from the participants was
that there were no rich or “better off” people in their community — all were poor or very
poor. This reaction came from anticipation of some form of. assistance. The principal
researcher and the discussant had to re-clarify the objectives one or two more times until
the ice was broken. In this study it was important to get the villagers’ assessment of the
key characteristics that constitutes a rich household, better off household, poor household

and very poor household.

Mamge System
The marriage system was included to get more information in support of a
question included in the questionnaire survey. The aim of this question was to probe the

vulnerability of households when one or both spouses die under matrilocal or patrilocal

263

marriages. CARE (2001), a non-govemmental organization (NGO) operating in Malawi
conducted a study and reported that women in patrilocal villages seem more vulnerable
because they have to leave husband’s home after the husband’s death. The security of
survivors seemed to depend on the type of marriage system one belonged to (whether

patrilocal or matrilocal).

Livelihood Activities

This topic was designed to get a consensus on the main activities and to obtain a
ranking of activities by their importance to household income. This will enable see
trends in priorities of households and household’s ability to cope with the shock of adult
mortality. There is a statement in Malawi that the North was blessed with education, the
Center with farming and the South with businesses. How true is this to rural households

in Malawi?

Constraints to Farming

Knowledge of constraints farmers are faced with was necessary to help design
interventions to improve agriculture. The aim of this question was to get a regional and
national picture of the constraints farmers faced in Malawi. It was important to assess if

farmers viewed chronic illnesses and death as constraints or not.

264

Chronic Illnesses, Death and Associated Losses to Households

This section solicited a listing and ranking of the main diseases and illnesses in
the communities as well as probe for the losses households were facing from these. This
section also established the coping mechanism at household and forms of assistance
coming the community, extended family, NGO and Government agencies to affected

households.

Small group Topics

Small Group discussions focused on asset depletion, knowledge of HIV/AIDS,
child-headed households and frequency of funerals. These topics were deemed sensitive

and were best handled in small groups.

Asset Depletion

The aim of this section was to find out how adult mortality has affected
households’ access to key assets such as land and livestock and to what extend
community mechanisms and institutions have helped surviving household members to
access the asset. Discussions ranged from what things are considered key assets to asset

liquidation during illness, to post death access and wife inheritance.

Knowledge of HIV/AIDS
The topic of HIV/AIDS was very important in the focus group discussions

because it allowed many issues to be explored. Discussions ranged from what HIV/AIDS

265

is to prevention, and sources where one can contract HIV/AIDS. One thing important to
note was people were very free to discuss the topic. It was encouraging to note that
within a space of two years only since the CARE (2001) study people had become more

open to talk about HIV/AIDS.

Child-Headed Families

Child-headed families are a new phenomenon emerging with the pandemic where
both parents had died of HIV/AIDS. This topic sought to establish if these h0useholds
were present in the surveyed communities and to find out how they were coping and their
sources of learning farming skills. The discussant and principal researcher had to make
an effort to narrow the discussion to these items. Most groups wanted to talk about AIDS
orphans. It turns out that many NGOs were registering orphans for assistance, hence the

tendency by groups to shift discussion towards AIDS orphans.

Frequency of Funerals

This topic asked the frequency of funeral during the past year and got an
assessment of the likely causes. Discussion was made on whether at funerals peOple

were announcing the cause of death.

266

Highlights From F GD Field Notes

The highlights from field notes collected during focus group discussions are
presented below beginning with Southern Region in section 3 of this report, then Central

Region in section 4 and lastly Northern Region in section 5.

3. SOUTHERN REGION FOCUS GROUP DISCUSSIONS

3.1 Background

Three Enumeration Areas in Blantyre ADD, Southern region had the highest
proportions of households with adult mortality. These were Kamwendo, Thumbwe and
Mbulumbuzi. Thumbwe had the highest with 8 out of the twenty households. Kamwendo
and Mbulumbuzi were tied at 7 each. Kamwendo EA was selected because it was
furthest away from both Thumbwe and Mbulumbuzi that are very near to each other and
to the city of Blantyre.

Thumbwe Enumeration Area is under Kadewere Traditional Authority in
Chiradzulu District. It is about 25 kilometers from Blantyre and enjoys heavily plied
roads that have minibuses and the Shire bus line. There is a tarmac road to the nearest
town (Boma) and most of the people travel to and from Blantyre daily to trade and work
in the city. The Kamwendo Enumeration Area is under Nkanda Traditional Area in
Mulanje District. It is nearly sixty kilometers away and has major transport problems.
Very few pick-up trucks ply the road and transport costs are exorbitant on this route. The

dirt road through Kamwendo is bad during both the dry and wet season.

267

3.2 Findings from Kamwendo and Thumbwe Enumeration Areas

3.2.1 Wealth Ranking

Table A3.1 below provides a summary of the characteristics of the each wealth
categories in Kamwendo and Thumbwe EAs. The last two categories, poor and very
poor are broadly similar in the Southern Region. Participants in Kamwendo insisted that
there were no farmers who belonged to the rich categories. In their view the rich were
those who had shops as well as rented extra land to cultivate. These people are usually
from outside the community who come to invest in trading business and then eventually
settled in the area and started farming as well. Because Thumbwe, is near the City of
Blantyre, the rich categories included people in businesses such as money lending, maize

milling, transportation and farming.

268

Table A3.]

Characteristics of the wealth ranking categories in Thumbwe and
Kamwendo numeration Areas, in Southern Region of Malawi

 

 

 

 

 

 

 

No change clothes

Washes clothes with maize husks
Begs for salt all the time
Chronically ill

 

Wealth Thumbwe EA Kamwendo EA
Categry
RICH Corrugated iron roofed house Owns a grocery / shop
Vehicle Good furniture (Sofas)
Employs labor on farming Well endowed children
Money lender Enough food
Sends children to school Has money in thousands
Enough food
Enough clothing
Livestock - cattle and goats
BETTER OFF Possess soap for washing and bathing Maize granary
Works in own garden Hires extra land to farm
Has money saved (around K2000) Has an orchard
Good clothing
Burnt brick house
Livestock — 2 cattle, chicken, guinea
fowls
Ox-cart
. Uses fertilizer
POOR He farms but does not harvest to last a No enough food to last entire year
whole year Relies on ganyu
Relies on piecework (ganyu) No domestic animals
Doesn’t use fertilizer on maize Doesn’t use fertilizer
Poor housing with no assets Grass thatched house with door made of
grass,
Normally dresses shabbily
VERY POOR Ever begging Lives in a very damaged house
No reliable accommodation Ever begging
No food No food

Engages in piece work

 

Source: Focus Group Discussion, MAPAMS 2002

3.2.2 Marriage System

The dominant marriage system in both EAs is matrilocal system (chikamwini). In

this system the man moves to stay at the wife’s family. No bride price is paid in the

system. If the wife’s family accepts any bride price, it is viewed as selling off a daughter

269

 

and it incurs the scorn of the village. In Thumbwe, participants mentioned that there is a
practice called “ulooka ” where a wife can go and stay with the husband for a number of
reasons. (i) If the woman’s home area has little land for cultivation, the family may
decide to move out, (ii) if there is bad attitude by the relative of the wife towards the
husband and (iii) due to selfishness. The children in Chikamwini and ulooka belong to
the wife. In the event of a divorce, the husband or wife must produce a token (Mzimba)
to signal the end of the marriage. This is usually a small fee like 10 Tambala or a piece
of cloth or a bark of a tree. In Kamwendo, the village headman, based on their
assessment of the man’s well being, now insisted that a divorcing man should build a
house for the upkeep of the wife and her kids before he could preside over the divorce or
pay some money. This point was hotly debated at the meeting because many viewed it as

a punishment, but the traditional authorities insisted that it was compensation.
3.2.3 Livelihood Activities

The main source of livelihood for the rural households in the Southern Region is
farming. The second source depended on proximity to a large trading center, town or
city. Table A3.2 below shows that Kamwendo, which is far away from Blantyre, ranked
contract on farming as the second main source of livelihood. Contract on farming is the
local term given to piecework during the dry season to prepare gardens. Land
preparation is done before the rains and is quite labor intensive. The average charge for
contract is K250/acre. During the season they do ganyu (casual labor) for weed control at

an average charge is 5 tambala per planting station per line or K270/acre in Kamwendo

270

and 10 tambala per planting station per line or K540/acre in Thumbwe. Brewing of both
the traditional opaque beer and Kachasu, an illegal spirit that has a high percentage of
alcohol and can causes liver sclerosis, is very common in both areas. Kachasu is.

Table A3.2 A Ranking of main sources of livelihood activities in the Thumbwe
and Kamwendo Enumeration Areas (EAS)

 

 

 

Ranking Thumbwe Kamwendo
First Farming Farming
Second Small Scale Business — Contract on Farming

Basket weaving, mandazi,
(flour fat-cooked buns)
beer brewing, opaque and
kachasu (illegal spirit)

 

Third Contract on farming Small-Scale Business —
Keeping livestock, selling
tomatoes, sugarcane, baking
cakes, mat making, pot making
and beer brewing

 

 

 

 

 

Source: Focus Group Discussion, MAPAMS 2002

3.2.4 Constraints to Farming

Several constraints to farming were identified in the focus group discussions in
the South. Constraints included soil fertility, poor weather, hunger and mortality. Table
A3.3 below gives a listing of these constraints for each of the EAs. Kamwendo EA
participants mentioned theft, poverty and hunger as problems that are unique to
Kamwendo compared with Thumbwe. Households did not harvest enough to last a
whole year. Most of the maize was consumed as green maize (Chitibu) because of the
hunger and fear of losing the crop to theft. Illness was mentioned as a constraint in both

EAs but death was only mentioned in Thumbwe.

271

Table A3.3 Lists of constraints to farming in Thumbwe and Kamwendo
Enumeration Areas

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Thumbwe EA Kamwendo EA
Poor soil fertility Soil water logging
No targeted inputs issued to farmers Lack of capital to buy fertilizer
Too little land Hunger
Chronic illnesses Lack of farming equipment
Poor weather- uneven rainfall distribution Lack of capital to buy Seeds
Poor market High fertilizer prices
Death among family members Theft
Chronic illness
Poverty
Lack of chemicals to put in
horticultural dimbas
Poor market and marketing

 

Source: Focus Group Discussion, MAPAMS 2002

3.2.5 Chronic Illness, Deaths and Associated Losses to Households

The major illnesses reported in Thumbwe and Kamwendo were malaria,
HIV/AIDS, diarrhea, malnutrition, asthma and cholera. A ranking of the main illnesses
that caused death in the areas is shown in Table A3.4 below. During the illness and
death, the participants at the Thumbwe FGD identified that they lost the skills of the sick
and deceased, labor and assets such as pots, sleeping mats, tins or pails taken to the
graveyard. Participants in Kamwendo mentioned loss of time and assets such as maize
and livestock to get cash to send the sick to hospital and to bury the deceased. In both
areas, households mentioned that the only coping mechanism available to them was to get

back to farming.

272

Table A3.4 A ranking of the main illnesses that cause death in Thumbwe and
Kamwendo Enumeration Areas

 

 

 

 

 

 

Ranking Thumbwe Kamwendo

First Malaria HIV/AIDS

Second Cholera Cholera

Third HIV/AIDS and ' Measles
Tuberculosis

Fourth Malnutrition Bloody diarrhea

Fifth Asthma Tuberculosis

 

 

 

 

 

Source: Focus Group Discussion, MAPAMS 2002

Help during illnesses and death came from several sources. Thumbwe mentioned
that they got financial support from extended families, church people assisted with food
and money and counseling, village headmen asked people to provide food, and the
government help was recognized as that of providing free drugs, transport to patients and
the deceased for burial. Kamwendo reported a very grim situation. The participants
reported that extended families did not help much while the community gave little
support in the form of 5 tambala per house. Because Kamwendo was far from Blantyre
and Zomba, Government and Non-Govemment were not forthcoming with any
assistance. The road networks are very poor in Kamwendo EA. When one had a dead
relative in Zomba hospital, they had to hire an expensive private vehicle to transport the
remains. There were reports that ambulance drivers sometimes dumped corpses and
accompanying relatives on the road to find their own means because the roads are bad.
These relatives then resorted to transport the deceased’s corpse by whatever means they

could get including bicycles.

273

3.2.6 Asset Depletion

Three key assets mentioned by men Southern region were food, bicycle and
livestock. Women considered children as the first key asset. The second and third assets
were different among women in the two EAs, for example, food came second in
Thumbwe, while in Kamwendo it was kitchen utensils that came second. Table A3.5 five
shows that there were marked differences between what men and women considered key

assets.

Table A3.5 Things considered key assets in Thumbwe and Kamwendo
Enumeration Areas

 

 

 

 

 

 

 

 

 

 

 

 

Thumbwe EA Kamwendo EA

Men’s Group Women’s Group Men Group Women’s Group
Maize — food Children Food Children
Bigycle Food, Bicycle Kitchen utensils
Livestock — goats, House Livestock money
chicken, cows '
Clothes, bedding Clothes Kitchen utensils Furniture
Farming equipment Beddings House Good beddings
Kitchen utensils Kitchen utensil Radio Clothes
Wife Farmland, fertilizer Hoes Latrine

and farming inputs
Furniture Pit latrine, bathroom Axes Rubbish pit

and rubbish pit
Radio Washing basket

 

 

 

 

 

Source: Focus Group Discussion, MAPAMS 2002

During illness and death both men and women’s small group mentioned that there
was no gender bias in asset liquidation. The assets liquidated depended on the nature of
illness and the financial need. Assets were sold to get cash to buy drugs, or traded in

exchange for labor if the illness was during the farming season. The assets that were

274

liquidated first were maize, clothes and livestock like goats and chicken. In Kamwendo
one of the male participant mentioned a bicycle and radio but and all the other men
disagreed. There seemed to be reluctance on the part of men to sale these because they
were regarded a man’s pride and tool of trade.

Post-death access to assets presented a challenge in the Southern Region. If a
woman died, the husband could stay for a week after burial and he then asked to be freed.
At this point, he was free to take some assets and leave some with the children. In
normal circumstances, the man took the bicycle, radio and some livestock. If there were
no children in the marriage he took everything. Some men noted rare cases, where a man
chose to leave all assets for the children. If a man died, and the man was living at the
wife’s village, assets were distributed between his surviving wife and relatives from both
sides (ankhoswe). There have been many cases where husband’s relatives confiscate all
assets and left the wife and children in instant poverty. In one particular case of Mrs.
Phiri (not her real name) relatives took away everything including the iron roof sheet
from the building even though she had her deceased husband’s will. Women in the focus
group discussion mentioned that the will is not well respected because relatives will say it
was the wife’s idea anyway.

If both parents died at the same time the appointed overseer relatives (ankhoswe)
would remain as caretakers of assets until the children were big enough to take over.
There have been cases where the ankhoswe started squandering the assets. Other
relatives approached the chiefs who then intervened.

Wife inheritance (Chokolo) is not practiced in the matrilocal communities. The

practice was last practiced in the 19305 and 405. Assets that were liquidated during

275

illnesses and funerals and post funerals were not recovered. Participants reported that the
decline in asset base is a permanent decline. The reason mentioned for the inability to
recover the asset base is that the death of an adult was associated with the loss of labor,
and money. But if a man died, it means loss of household leadership to plan and recover

assets.

3.2.7 Knowledge of HIV/AIDS

There was a good knowledge of what HIV/AIDS is and the difference between
HIV and AIDS among the younger man than older men in the South. The same trend
was observed among the women’s group. There seemed to be a sense of comfort among
the elder generation that HIV IAIDS is a disease that would not catch them since they
were not as sexually active as the young people. They reported that they came to know
about HIV/AIDS though radio, clinics and drama groups. Women’s group added
churches, hospitals and other people from town. The common symptoms were reported to
be weakness, anemic, very thin, appetite for delicious meals e. g. fried meat, chronic
coughing, pale hair, persistent diarrhea, swellings and very aggressive. The participants
knew that one could only know his or her HIV status through blood test. Preventative
measures were pointed out as abstinence, faithfulness to one partner and use of condom.
Condom usage brought a heated discussion where the older people argued that condom
always bursts and had become the cause of the spread of HIV/AIDS because many
people took false comfort in the condom and stopped using self-restraint. If a woman got

condoms from the clinic and asked the man to use them, the men reported that they

276

would divorce her. It is considered an insult and the men claimed they don’t enjoy sex in
a condom. The elderly women stated that condoms encourage prostitution, if there were
no condoms, people would be faithful to one partner.

Men’s groups mentioned that the popular places where one could contract AIDS
were around the communities with taverns and drinking places and where prostitutes
solicited for sex from the drunk men, in rest houses, at local weddings and initiation
ceremonies done at night. Men also castigated women who went to town for small-scale
business and then get offered more money if they had sex with the rich men in town. The
women pointed out house taverns, trading centers, as well as initiation ceremonies done
at night. The women claimed that men with lots of money came around periodically and

encourage girls to “taste” sex (kusasa fumbi).

3.2.8 Child Headed Households

Participants reported the existence of child-headed’families with household heads
aged 12 - 16 years old and children ranging from 6 years old. These families rely on
piecework. Girls tend to marry early to escape the misery or are at risk of getting into
prostitution. Focus group participants reported that child-headed households do not get
community assistance and they learn farming methods from relatives through observing

other farmers.

277

3.2.9 Frequency of Funerals

In the Southern Region, participants reported that funerals rarely occurred during
the rain season. The frequency of funerals was 2 per week and 3-6 per week for a
community of 4 villages in the dry season. Funerals peak during the months of April,

May and June when there is food shortage.

4. CENTRAL REGION FOCUS GROUP DISCUSSIONS

4.1 Background

Selection of the Enumeration Areas in Kasungu ADD, Central Region differed
from the process in the Southern Region. Two of the highest adult mortality areas were
Lukwa, near the town of Kasungu and Chakaza near a large trading center of Mponela.
Lukwa and Chakaza had both 8 out of two households affected by adult mortality. It was
decided to select the northern most EA which is Tumbuka speaking and the southernmost
EA which is Chewa speaking near the border town of Mchinji. The Central region is
unique in that it has both languages as well as both marriage systems. Focusing on the
EAs with the highest adult mortality would have resulted in under representation of this
fact.

Kamwendo EA is near the first large trading center near the border town of
Mchinji on the Zambia/Malawi boarder. Kamwendo EA is in Mchinji District in
Traditional Authority Zulu. The area is a good tobacco producing area. Kaluluma EA on

the other hand is near to the northern border of the Central province. It is in Kasungu

278

District under Traditional Authority Kaluluma. The sitting Member of Parliament is from
the opposition whereas for Kamwendo EA the sitting MP is from the ruling party.

Both EAs are close to main tarmac roads and transport is not a major problem.
Kaluluma reported 2 deaths out of sixteen households while Kamwendo had 5 out of the

nineteen households interviewed.

4.2 Findings from Kaluluma and Kamwendo EAs

4.2.1 Wealth Ranking

Table A3.6 below provides a summary of the characteristics of the wealth
categories in Kaluluma and Kamwendo EAs. The better off, poor and very poor
categories are broadly similar in the Southern Region. A clear distinction is notable
between the rich settler farmers surround Kamwendo area while Kaluluma is quite far
from those. Participants in Kamwendo clearly describe characteristics of settler farmers
as those belonging to the rich categories. In Kaluluma, the rich were those households

with enough food to eat all year round and had good housing.

279

Table A3.6 Characteristics of the wealth ranking categories in Kamwendo and

Kavuta Enumeration Areas, in Central Region of Malawi

 

Enumeration Area

 

 

 

 

 

 

Wealth Kamwendo Kavuta
Category
RICH Livestock — cattle and goats Harvest enough maize to eat all year (6-
Bank accounts 13 oxcart full) .
Dimba Good house with burnt bricks & iron
Good house sheets
10 hectares of leased land Dresses well
Car Livestock (cattle, goats)
Stereos and video
Servants
BETTER OFF Bicycle Harvest 2 -3 oxcarts of maize
Radio Dresses well
Ox cart Livestock (cattle, goats)
Harvest enough food to last a Burnt brick house with thatch roof
year Buys one or two bags of fertilizer
Livestock (goats, chicken, doves) Bicycle
POOR Farms with no fertilizer No blanket
Has many children Insufficient food
Not enough clothes Does piece work
Shortage of food Dresses poorly
Depends on Ganyu for money May have a few chicken
VERY POOR Ever begging Sleeps on sacks
Inherit vacant houses Harvests nothing
No beddings No money, clothes, livestock

 

Sometimes mentally disturbed

 

Have many children

 

Source: Focus Group Discussion, MAPAMS 2002

4.2.2 Marriage System

The Central Region has two marriage systems. The matrilocal system is

predominant in the southern parts of the region and is the same system as that in the

Southern Region. In the Mchinji area, Kamwendo EA, the man is required to pay some

bride price ranging from 400 Kwacha for a young unmarried woman to 100 Kwacha if

the woman had been married before. In Kaluluma, the predominant system is more

patrilocal. Lobola or bride wealth is paid either as cattle (2 — 3) or the cash equivalent

280

 

(MK8,000 — 15,000) depending on the going market price of each cattle which normally

range from MK4,000 -— 5,000.

4.2.3 Livelihood Activities

The main source of livelihood for the rural households in the Central Region is
farming. Unlike in the Southern Region, farming is the mainstay and is thus subdivided
into garden farming where crops like maize and tobacco are grown, dimba farming where
vegetables are grown and then livestock farming. Population pressure in the central
region is less than that of the Southern Region and thus permits livestock farming. These
when ranked separately showed garden farming as the most important for Kamwendo
EA, followed by dimba gardens and then livestock as shown in Table A3.7 below. In the
northern most EA of Kaluluma, livestock farming and piecework ranked second and

third, after garden farming respectively. Small businesses ranked fourth in both areas.

Table A3.7 A ranking of main sources of livelihood activities in the Kamwendo
and Kaluluma Enumeration Areas (EAS)

 

Enumeration Areas

 

 

 

 

 

 

 

 

Ranking Kamwendo Kaluluma
First Garden farming Garden farming
Second Dimba farming Livestock farming
Third Livestock farming Piece work
Fourth Small Scale Business — Basket Small-Scale Business —
weaving, mandazi, beer Keeping livestock, selling
brewing, opaque and kachasu tomatoes, sugarcane, baking
cakes, mat making, pot
making and beer brewing

 

 

Source: Focus Group Discussion, MAPAMS 2002

281

4.2.4 Constraints to Farming

Table A3.8 lists the constraints to farming listed by the FGD participants. Lack of
capital to buy seed and fertilizer and chronic illness were common to both EAs. Issues to
do with pricing of both inputs and outputs featured among participants in Kaluluma.
Participant complained that input prices were higher and the price they received for their
produce net transport charges was low. There was an ADMARC sub—depot at the local
trading center that was closed due to some structural adjustment. Since its closure,
farmers now have to go to Nkhamenya to buy and sell hence the resulting less favorable

prices. Hunger as problem, was identified as a Kamwendo.

Table A3.8 Lists of constraints to farming in Kamwendo and Kaluluma
Enumeration Areas

 

 

 

 

 

 

 

 

Enumeration Area
Kamwendo Kaluluma
Lack of seed Low rainfall
Lack of fertilizer Lack of capital to buy fertilizer and seed
Hunger Low price of produce
Chronic illnesses Chronic illness
Termites High fertilizer prices

Witch weed

 

 

 

 

Source: Focus Group Discussion, MAPAMS 2002

4.2.5 Chronic Illness, Deaths and Associated Losses to Households

The major illness reported in Kamwendo and Kaluluma were malaria, HIV/AIDS,
diarrhea, malnutrition, chronic coughing, scabies and cholera. A ranking of the main

illnesses that caused death in the areas is shown in Table A3.9 below. During the illness

282

and death the participants at both Kamwendo and Kaluluma mentioned that they lost
labor, cash and assets through liquidation to raise money for drugs or hospitalization of a
relative. The loss of skill was mentioned at Kamwendo while in Kaluluma, they pointed
out that there would be inadequate supervision of gardens. In Kamwendo, participants
reported that households that are affiliated with religious organization received assistance
when the local parish organized help to work in the field. For funerals, the local village
collected cash and maize flour to help feed the mourners. In Kaluluma, the participants
mentioned that family members helped by providing financial assistance and support. An
NGO, Plan lntemational, had also helped by initiating and facilitating a drug revolving

fund to assist the community.

Table A3.9 A ranking of the main illnesses that cause death in Kamwendo and
Kaluluma Enumeration Areas

 

 

 

 

 

 

 

 

 

Enumeration Areas
Ranking Kamwendo Kaluluma
First Malaria Chronic coughing
Second Chronic Coughing Diarrhea
Third Scabies Malaria
Fourth Hunger & Malnutrition HIV/AIDS
Fifth HIV/AIDS STD — gonorrhea, syphilis

 

 

 

Source: Focus Group Discussion, MAPAMS 2002

In both areas, government support was listed as transportation of the sick to hospital and
delivery of the corpse if the person is deceased. In Kaluluma, one man was quick to point
out that the government only carried back the body of the person they brought to hospital

in an ambulance.

283

4.2.6 Asset Depletion

Cattle, clothing, money, food and land are considered key assets in Central
Region. The women’s group mentioned livestock to include not just cattle but goats,
chickens and ducks. A striking difference between man and women’s group is that men
were silent about money as an asset while women begged to differ. Men usually spend
money drinking and hence did not consider it an asset. Table A3. 10 highlights what men

and women considered being key assets.

Table A3.10 Things considered key assets in Kamwendo and Kaluluma
Enumeration Areas

 

 

 

 

 

 

 

 

 

 

 

 

 

Kamwendo Kaluluma
Men’s Group Women’s Group Men Group Women’s Group
House Shelter Livestock (cattle, Money

goats, poultry)

Livestock Food Oxcart Food
Oxcart Money Hoes ' Children
Radio Clothes, House Chairs
Land Kitchen Utensils Big land Kitchen utensils
Marriage Livestock Children Land
Food Furniture Furniture Livestock
Woodlot Farming land
Bicycle Children

 

 

 

 

Source: Focus Group Discussion, MAPAMS 2002

As is the case with the other regions, both men’s and women’s small groups did
not show gender bias in asset liquidation to meet cash or food expenses during illness or
death of a loved one. The assets were liquidated based on the nature of illness and the
anticipated expenses. The assets that were liquidated first were maize, clothes, goats and

chicken.

284

Post-death access to assets presented less of a challenge in the Central Region
than in the South. When a man died in Kaluluma area, a brother could inherit his late
brother’s wife — Chokolo. He took over everything including the welfare of the surviving
spouse and children. The woman was free to decline inheritance. If she declined to be
inherited, she could remain in-charge of the family assets as long as she did not remarry.
If she wanted to remarry she had to depart leave all the assets for the children. She was
only allowed to take a few assets if she were caring for very young children that needed
maternal support. Wife inheritance is still strong among the Tumbuka speakers. If a wife
died, the surviving husband is allowed to remarry after 6 months to a year.

Wife inheritance is not practiced in the matrilocal communities in the Central
Region just as in the Southern Region. Assets that are liquidated during illnesses and
funerals and post funerals were not recovered. Participants reported that permanent
decline in asset base was normal. In a few cases, relatives of the deceased contributed
money and gave the surviving members some money to start some income generating

projects.

4.2.7 Knowledge of HIV/AIDS

There is a good knowledge base among the community on what is HIV/AIDS
among men and women in the Central Region. They reported that they came to know
about HIV/AIDS though radio, clinics, drama groups and seeing sick relatives. The
common symptoms were reported to be pale hair, coughing, and loss of weight, persistent

diarrhea, brain disturbance and lack of appetite. The participants knew that one could

285

only know his or her HIV status through blood test. Preventative measures were pointed
out as abstinence, faithfulness to one partner and use of condom. Condom issue
generated much discussion. The older man argued that the condom breaks and had been
the cause of more HIV IAIDS infection. One old man mentioned that they opened a
condom and displayed it in the sun and saw some microbes moving about in the
lubricant. He was convinced that the lubricant cam'es some viruses that could themselves
cause AIDS. A young man produced a condom and challenged the old man to prove the
point. Pressed further, the older man argued that having sex with a condom is like eating
a candy in a plastic cover defeating the point of eating the candy in the first place. The
elderly women stated that condoms encourage prostitution, if there were no condom
peOple would be faithful to one partner.

The men’s groups pointed out that AIDS could only be contracted in major towns
such as Lilongwe, Blantyre, Mzuzu and border towns like Mwanza and Mchinji. They
viewed Nkhamenya and local trading centers to be safe. They blamed city men who
came out to pollute their local prostitutes since they could pay more money. The
women’s group mentioned local trading centers and taverns. They also disclosed that
because of poverty and hunger that was prevalent this year some women were deceived
by the lure of money and traded sexual favors for money to buy food and got infected.
Both groups admitted that the practice of multiple sexual partners is prevalent in the
community. However, it was encouraging to note that both groups talked to their family

members about the dangers of AIDS.

286

4.2.8 Child-headed Families

Child headed families were reported as existing in the community. They survive
by getting employed to take care of livestock, work at the market or by begging at trading
centers. They learned farming methods from fellow villagers. In Kamwendo, the
community helped by facilitating counseling and in limited cases by adopting young

children.
4.2.9 Frequency of Funerals

Funerals frequency was reported high due to the hunger situation. Those who died
were suffering from malnutrition and had swollen bodies. The general trend had been
high death rate during hunger and floods in the rainy season, and low rate during the dry
season following harvest. Participants reported up to 2 funerals per week in a community

of four villages.

5. NORTHERN REGION FOCUS GROUP DISCUSSIONS

5.1 Background

Three enumeration areas in Mzuzu ADD, Mpherembe, Kamchocho and
Mzalangwe each had seven households each affected by adult mortality. Kamchocho and
Mzalangwe were randomly selected out of those three. Kachocho EA is in the central

part of Mzimba District while Mzalangwe is in the northern part of Mzimba. Mzimba

287

District is the largest district in Malawi. It has potential to be subdivided into three
districts. The proposal to subdivide it has been rejected several times by the Traditional
Authority of the area who sees the move as a division among the An goni People.
Unfortunately, it is a disadvantage to the constituency in that the whole district will
receive one hospital, one program and one of everything allocated per district by the
government instead of receiving three such items.

The road networks are poor in the district and agricultural and other government
departments are overstretched to cover the area. The area is predominantly Tumbuka
speaking and the marriage system patrilocal. Mzimba and much of the north is mainly a

domain of the opposition party.

5.2 Findings from Kamchocho and Mzalangwe Enumeration Areas

5.2.1 Wealth Ranking

Table A3.11 below provides a summary of the characteristics of the wealth
categories in Kamchocho and Mzalangwe EAs. Kamchocho, like Kamwendo in the
Central Region was nearer settler farms. Participants viewed those settler farmers as the
rich. In Mzalangwe the rich are those that harvest enough food to last a year. Participants
strongly pointed out that the wealth of a household is most seen in its ability to produce a
good maize harvest enough to feed the family all year round. Number of livestock was
the next important criteria. Unlike in Southern and Central Region, the Northern region

used cattle for land preparation. Livestock was therefore both an important asset farming

288

asset and a form of saving. The better off were those households that had anywhere up to

10 cattle. The poor and very poor categories were broadly similar to those in other

regions.

Table A3.11 Characteristics of the wealth ranking categories in Kamchocho
and Mzalangwe EAs, in Northern Region of Malawi

 

 

 

 

 

 

 

 

 

Wealth Enumeration Areas

Category Kamchocho Mzalatflw—e

RICH Enough Food Harvest enough maize for food
Vehicle Uses fertilizer when farming
Estates Livestock (cattle, goats)
Money in the Bank Good housing
Livestock Has land
Hired laborers Dogs for security

BETTER OFF Use fertilizer Can afford 1 bag of fertilizer
Harvest enough food to last a Dresses well
year Livestock (cattle, goats)
5 — 10 Cattle Burnt brick house with thatch
Other livestock (goats, chicken, roof
doves) Harvest food that last 50-75% of
Good housing the year

POOR Ever begging No cattle
Does piece work (ganyu) No livestock
Eats one meal a day Little land
Poor housing Little meal per day

VERY POOR No livestock ever begging
No food No housing
Dresses shabbily No money, clothes, livestock
No desire to have things Borrows clothes

 

Source: Focus Group Discussion, MAPAMS 2002

5.2.2 Marriage System

The marriage system in the Northern Region was reported to be predominantly

patrilocal. Lobola or bride wealth is paid to the in-laws in the form of cattle. For a young

unmarried woman, the bride price was 4 cattle; l steer, 1 heifer and 2 cows. The

289

 

monetary value of the bride price ranged from 20,000 - 40, 000 Kwacha depending on the
going price of cattle in the area. The price per beast seemed to increase as you go further
north. If the woman had been married before or had a child outside marriage then the
price fell to 2,000 — 3,000 Kwacha. Participants reported that marriage between sons
and daughters from the North and spouses from the South were increasing and were
acceptable. When men from the South married their daughters, lobola had to be paid
and when, sons from the north marry women from the South and no lobola was paid, the

children born belonged to the patrilocal inheritance practices.

5.2.3 Livelihood Activities

Garden farming was the main source of livelihood for the rural households in the
Northern Region. Dimba gardens were the second most important livelihood source in
Kamchocho. In Mzalangwe participant reported that the Dimbas in the area have long
since dried up. These only operate in good rainfall years. Piecework was more important
in Mzalangwe. Participants in Kamchocho reported that piecework was more prominent
in good years. The going rate for land preparation piecework charge was around 800

Kwacha per acre. Small scale business ranked third as shown in Table A3. 12 below.

290

Table A3.12 A Ranking of main sources of livelihood activities in the
Kamchocho and Mzalangwe Enumeration Areas (EAS)

 

 

 

 

 

 

 

Enumeration Areas

Ranking Kamchocho Mzalangwe

First Garden farming Garden farming

Second Dimba farming ' Piece work

Third Small Scale Business — Small-Scale Business —,
buying and selling fish, selling tomatoes, baking
mandazi, beer brewing, cakes, house and pit latrine
butchery and vegetable making, and beer brewing
produce

 

 

 

Source: Focus Group Discussion, MAPAMS 2002

5.2.4 Constraints to Farming

Table A3. 13 shows the farming constraints listed by the FGD participants. High
input price and lack of markets were common to both EAs. Closure of ADMARC rural
marketing facilities affected the Northern Region adversely because of the vastness of the
area and the low population densities. Farmers were now facing high input prices and low
produce price due to high transportation cost. Erratic rainfall in Mzalangwe affected
yield levels and also caused dimba areas to dry up. In 2002 season most participants in
Mzalangwe reported a poor harvest insufficient to last a whole year. Poor road
infrastructure was also a major constraint in both EAs. Transportation was provided
private pick-ups (Matolas) who charge exorbitant prices. Minibus operators do not want
to ply those routes due to the poor condition of the roads. Participants in Mzalangwe EA

also mentioned the problem of New castle disease in chickens.

291

Table A3.13 List of constraints to farming in Kamchocho and Mzalangwe
Enumeration Areas

 

 

 

 

 

 

 

 

 

 

 

Kamchocho EA Mzalangwe EA
Shortage of draft animals Poor soil fertility
Chronic illness/death Erratic rainfall
High fertilizer prices No Water for dimba cultivation
Shortage of cash to buy fertilizer Poverty
No markets High fertilizer and seed prices
Collateral problems to get fertilizer New castle disease with chicken
Poor roads Low crop prices
Irnpassable roads to transport inputs and
produce
No markets
Late distribution of TIP

 

 

 

 

Source: Focus Group Discussion, MAPAMS 2002

5.2.5 Chronic Illness, Deaths and Associated Losses to Households

Malaria, HIV/AIDS, diarrhea, malnutrition, chronic coughing, scabies and cholera
were reported to be the major diseases prevalent in both Kamchocho and Mzalangwe
Enumeration Areas. A ranking of the main illnesses that caused death in the areas is
shown in Table A3.14 below. Some participants at both FGDs reported that they lost
labor and cash during illness and death of a relative. Loss of skill and knowledge were
not mentioned. In Kamchocho a point was raised that school children lost schooling days.

Participants in Kamchocho painted a rather bleak picture when they reported that
no help was forthcoming from community, extended family members and NGOs. They
did point out that the hospital provided transport for the remains of the deceased. In
Mzalangwe, participants reported that local villagers collected cash and mealie-meal to

help feed mourners at funerals. Extended families contributed money to help care for the

292

sick and assist farming in their relative’s field while they were away at the hospital. Both

FGDs reported that government helps providing transport for the dead body.

 

 

 

 

 

 

Table A3.14 A ranking of the main illnesses that cause death in Kamchocho
and Mzalangwe Enumeration Areas
Ranking Kamchocho Mzalangwe
First Malnutrition HIV/AIDS
Second Cholera Chronic Coughing
Third Malaria Malaria
Fourth Scabies Diarrhea
Fifth Cholera

 

 

 

 

Source: Focus Group Discussion, MAPAMS 2002

5.2.6 Asset Depletion

Table A3.15 highlights what men and women considered key assets. Children are

considered as assets in patrilocal system. The main gender difference noted was that

women considered kitchen utensils as assets while men did not.

 

 

 

 

 

 

 

 

 

 

 

Table A3.15 Things considered key assets in Kamchocho and Mzalangwe
Enumeration Ereas
Kamchocho EA Mzalan e EA
Men’s Group Women’s Group Men’s Group Women’s Group
Cattle Food Children Land
Ox-cart Bedding Clothes Kitchen Utensils
Bicycle Clothes Livestock House
Goat Kitchen utensils Food Dogs
Furniture Furniture Furniture Children
Food Mat Livestock
Children Pails Furniture
Money Livestock Maize
House Oxcart

 

 

 

 

 

Source: Focus Group Discussion, MAPAMS 2002

293

 

 

As is the case with the Southern and Central regions, both men’s and women’s
small groups mentioned that there was no gender bias in terms of which assets were
liquidated to meet cash or food expenses during illness or death of a loved one. The
assets were liquidated based on the nature of illness and the anticipated financial cost.
The women’s small group however pointed out that if the woman was sick and
hospitalized away, there was a rather wanton sale of assets in the name of raising money
for the sick. In most cases, they hid important assets so that they don’t get sold. Maize,
clothes, goats and chicken were liquidated first.

Post-death access to assets was more assured in the northern region because of the
patrilocal system. The clothes of the deceased were the only things that were shared
among relatives of the deceased. Land and other assets were kept for the surviving
members. FGD participants reported that relatives from the husband’s side were normally
appointed to take responsibility for the assets until the kids were grown ups, at which
point they would redistribute the assets among the children. If there was an elder son in
the family, he was given that responsibility. Wife inheritance was still being practiced in
the Northern Region. Although the elders were now encouraging both the wife and the
deceased’s brother to go for HIV/AIDS testing before the inheritance ceremony is
performed. The woman was free to decline inheritance. If she does so, she could remain
in-charge of the family assets as long as she did not remarry. If she opted to remarry she
had to depart, and leave all the assets for the children.

The participants reported that there was cases where appointed relatives took
advantage of the young children and abused the assets by selling them in the name of the

kids but then benefited themselves. In such instances the village headman Would call

294

other relatives and resolve the matter. Assets that were liquidated during illnesses and
funerals and post funerals ceremonies were not recovered. Participants reported that the

decline in asset base was rather permanent than temporal.

5.2.7 Knowledge of HIV/AIDS

There was very strong knowledge of HIV/AIDS in the Northern region despite the
remoteness of some of the areas. Participants reported that they came to know about
HIV/AIDS through radio, clinics, and drama groups. The common symptoms were
reported to be pale face, shingles, red lips showing sores in mouth, weakness, fever,
diarrhea, persistent skin rashes, anemic, appetite for meat and loss of weight. The
participants knew that one could only know his or her HIV status through blood test.
Preventative measures were pointed out as abstinence, faithfulness to one partner and use
of condom.

As in the Southern and Central Region, the condom issue generated much
discussion with older man and women arguing that the false security given by condoms
has been the reason why HIV/AIDS has spread across Malawi. It was apparent from the
sessions that young people were using condoms whereas the old people did not. One
elder man stated that he did not pay lobola in-order to use a condom with his wife.

The places where one can contract AIDS were pointed out to be boarding schools,
trading center taverns and in the bushes as well as when getting treatment from traditional
healers who use razor blades to make cuts to administer traditional medicines. The

women’s group in Mzalangwe pointed out that boarding school kids are usually confined

295

at school during school terms were most at risk, as they tend to have unprotected sex
whenever they get the slightest opportunity during school breaks. During market days,
people come from Mzuzu with lots of money and merchandise to sell. Those men lure
young girls and some married women into sleeping with them for money and hence they
end up getting HIV/AIDS. Participants reported that husbands do speak to their sons
about AIDS while wives speak to their daughters about AIDS. It is difficult for wives to

speak to sons and fathers to daughters.

5.2.8 Child headed Households

Participants reported that there were a few child headed households in both
Kamchocho and Mzalangwe EAs. These child headed households were surviving by
doing ganyu in the villages. They received farming mentorship close relatives.
Community and NGO assistance was generally reported lacking. Focus group
participants had only heard of NGO helping out in areas nearer to the road. Compared to
the other regions, kinship was strong in the north such that these households got help

from the social capital surrounding them.

5.2.9 Frequency of Funerals

Funerals were reported to be quite frequent at about 3 — 4 per week in a

community of four or five villages. A few of the people dying were sons and relatives

who were working outside the community, either in towns or in other countries like

296

South Africa. The participants mentioned that local villagers were mostly dying from

hunger and malnutrition.

6. EMERGING ISSUES

A number of issues emerged from the focus group discussion exercise pertaining
to the Malawi Agricultural Productivity and Adult Mortality Survey project, the donor
community, Government of Malawi and communities. These are highlighted in sections

below.

MAPAMS Project

Results of the focus group discussion called for caution to be exercised in the
analysis of asset data through recognition of the fact thatthe rich and better off categories
of wealth ranking were localized. The marital system seemed to have a very strong
bearing on security of household in as far as post death access to assets was concerned.
There seemed to be more security for survivors in the patrilocal than matrilocal system.
Design and implementation of mitigation measures should take this into consideration.
Marital system had a bearing on existence and the level of safety nets available to child
headed households as well.

HIV/AIDS was no longer a very sensitive topic as it was during the previous 2 to

3 years. There was more openness in talking about HIV/AIDS. The level of openness

297

was deemed commendable, as it would shape future projects to mitigate negative AIDS

impacts.

Donor Community

There were two areas identified to be potential areas for interventions by the
donor community based our focus group discussions. First, the FGDs confirmed the
existence of child-headed households, though exercises could not ascertain the level of
prevalence in communities. There was a need for a more focused study that would
quantify these households and form the basis for NGO programs that cater for not just the
welfare of these child-headed households but agricultural knowledge and skills training
and technology transfer, HIV/AIDS awareness and ensuring that surviving children got
access to education.

Second, the reluctance to use condoms and misconstrued beliefs on condoms and
condom use among the elderly was worrying and could leave them vulnerable. There
seemed to be a resistance among the older people to use condoms or endorse their use,
based on all sorts of excuses. We identify a need for NGOs to target this group and come

up with means and information that allay their fears of condom breakage and other

excuses. The elderly need to be mobilized to be allies in the war on HIV/AIDS.

298

Government of Malawi

Some issues emerging from the FGD needed the attention of the Government of
Malawi. In the Northern Region, and Mzimba District in particular, agricultural
productivity in addition to being affected by adult mortality was being severely hampered
by poor infrastructure. There was a need to improve road networks in that area so as to
lower transactions cost for rural farming households. Tied to the infrastructure issue, was
timeliness of agricultural inputs. Farmers reported getting Targeted Input Program (TIP)
inputs late in January. There was therefore a need to ensure not just TIP inputs but also
other agricultural inputs are available to farmers on time.

It appeared that hunger caused more deaths in some parts of rural Malawi than
what was reported in the media. Most participants reported that they did not harvest
enough maize grain in 2002 because of drought, theft, and consumption of maize at the
green stage. There was therefore a need for the government to step up on-going effort to

stock up grain acquisition to fill and maintain strategic grain reserves.

Local Communities

There was a need for local communities to organize more HIV/AIDS awareness
campaigns at community level to keep the issue of HIV/AIDS topical especially among
the elderly people and not just the young people. Community leaders needed to mobilize
their members to do more philanthropic acts to help households affected by adult

morbidity and mortality. Lastly, there was a need for religious organizations that were

299

currently actively involved for their members only to reach beyond congregations and

complement other efforts by community based organizations and NGOs.

7. CONCLUSION

The focus group discussions were quite instrumental in exposing issues that
would otherwise not be captured using formal survey. Among these were the collective
construction of wealth ranking categories, the insecurity of survivors in the marriage
systems and attitude of men towards use of condoms. The FGD had the main advantage
of flexibility to pursue emerging issues coming from a lively discussion. Respondents
were quite open and very informative. The results from the FGD have produced
intermediate preliminary recommendations to the MAPAMS project, donor community,
Government of Malawi and local communities, but most importantly helped to fine—tune

the MAPAMS questionnaires.

300

References

CARE. 2001. The impact of HIV/AIDS on Agricultural Production Systems and Rural
Livelihoods in the Central Region of Malawi. Report for USAID/Malawi,
Lilongwe.

301

APPENDIX B

Appendix B: HIV/ADS Pathways of Influencing Rural Households

302

.ooom .20am...0 Na? 9.35 new :oEmm ”mmmw O<n_ ”E9: USQMUd.

 

 

was 85:52.: 65:8 2.2 £2838 38:3 EoEEo>om 4620 8 388 ”£968 3:25:35 o conu— 98 EE— uo 32988 3:22
.59: £8.35 ”890$ oEocoom o 8886 638:0 ”9888 Bofibfi 3m 0 2089293 Baogmacfi ..mmooom .819: ”83:28qu REES—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

«uanam wagon: 2393 .350

1 4 ll...
..lv 5:923 . . ........... 2:85 ..................u
. 50:: new. 5 a I/ . 2.38am? c. e ..... .. u
n F k I, u . . .
u < n u
. 1 J .coaoauoa x0232.
”i .v 820 @2225 . - .. . new no.0 c. a ..... cougfifl m3 Boon .832 85582
u .30. 8 £5 . 335 ES. 89.23 c. ._, ..mofioE c. e .0 $0..
. f L .
n 4 u s m
u g 8.3.5.5 53.2 .38 a u n 3
m 20.5.0300"... ESE mm_._._>.._.o< Scenic m m
m a .33 A. m m
u " noon. " u u
n u 2.3 Allis " n n
u n i 30:33: 25:. imam " H ”

£923 226 a x06 .2 98 o. Loam. .0 £933 223 w x07. ..2 9.8 2 .83
56520 scam. 9:02.65 co $3 .0 56.96 goon. 626336 6 $0..
... t s
m ................ . EEQE I 9.23 <n_ $0ch 330% 5 98952 . ................ m

.305 330% .0 use... no 58236:...

ll

Alli 1‘7
0.2moz<n_ wan—(3....

 

 

 

 

«20:330.. .55 .83 Bantu 3283:. new 3.29.2: ammonia mag—(3.... p.023 .3 $6353 .2228 .5 0.59“.

APPENDIX C

Appendix C: Model Selection - Double Hurdle vs Tobit and Biprobit vs Probit

304

Table Cl Determinants of participation in off-farm income sector by adult individuals in rural

households in Malawi, 2002 MAPAMS study. Unrestricted Double Hurdle Model

estimation (with PA Death Variables)

Dependent variable: Participation ( l=if individual participates; 0=otherwise)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Double Hurdle Model
Independent Variables Estimate I Standard Error | Marginal Effect
Individual Characteristics ~
_Age 0.1 186*** 0.0224 0.0316
_Age squared -0.0013*** 0.0003 -0.0004
Education -0.0040 0.0199 -0.0011
Female -0.1757 0.1455 -0.0469
Household head 0.4571 0.5916 0.1326
Household spouse -0.0352 0.2012 -0.0093
Adult -0.1 122*** 0.0332 -0.0299
Household Characteristics
Dependency ratio -0.0524 0.0597 -0.0140
Female headed 0.0727 0.2348 0.0199
Age of head 0.0027 0.0117 0.0007
Education of head 0.0376 0.0328 0.0100
Housing Value (MK) -2.8e-06 2.22e-06 -7.47e-07
Landholding 0.0056 0.0485 0.0015
Grow Cassava before 2002 -0. 1963 0.1828 -0.0486
Club membership -0.2751* 0.1644 -0.0664
Location-access characteristics
Distance to bus stop 0.0034 0.0124 0.0009
Distance to market -0.0076 0.0104 -0.0021
Kasugu ADD 0.0413 0.1138 0.0110
Blantyre ADD -0.361 1** 0.1472 -0.0900
Death Variables (with Timing 0 years afier a PA death)
PA male hos death 0.3636** 0.1749 0.1 101
PA female hos death -0.1561 0.2226 -0.0390
PA male nhos death 0.1888 0.2034 0.0541
PA female nhos death 0.2828 0.1906 0.0837
Constant -2.4222 0.4513 -
LR Chi Sq Statistic 222.85
p—value 0.0000
R-squared 0.2222
Log likelihood -458.04
N 1075

 

Source: MAPAMS 2002. *** 1% Significance, **5% significance, *10% significance

Notes
hos = female head or spouse,

nhos = non head

305

 

Table C2: Determinants of participation Intensity in rural non farm employment by adult

individuals in rural households in Malawi, 2002 MAPAMS study. Restricted Double
Hurdle and Tobit Model estimation results (with PA Death Variables)

Dependent variable: Off-farm work time (number of hrs worked per week)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Double Hurdle Model Tobit Model
Independent Variables Estimate Standard Z Estimate Standard 2
Error Statistic Error Statistic
Individual Characteristics
_Age 0.7699 0.4878 1.58 0.71 14 0.4958 1.43
_Age squared -0.0056 -0.0057 -1.15 -0.0067 0.0059 -1.13
Education 0.4400 0.5136 0.86 0.3925 0.5004 0.78
Female 4.6371 4.3855 1.06 3.4802 3.9467 0.88
Household head 13.9886 10.746 1.30 12.1074 11.557 1.05
Household spouse 4.5383 3.4086 -1.33 -4.6923 4.7000 -1.00
Adult -0. 1402 0.7875 -0. 18 -0.0851 0.9094 -0.09
Household Characteristics
Dependency ratio 1.8575* 1.1385 1.63 2.3972* 1.3538 1.77
Female headed -5.9688* 3.1867 -l.87 -6.4423 5.0564 -1.27
_Age of head -0.2364 0.2204 -1.07 -0.1989 0.2341 —0.85
Education of head 0.1 178 0.6054 0.19 0.4100 0.6504 0.63
Housing Value (MK) 1.89e-06 0.0001 0.04 -1.67e-06 0.0001 003
Land holding (ha) 1.6672* 0.9799 1.70 1.9574* 1.1605 1.69
Grow Cassava before 2002 -2.6493 5.9492 -0.90 -3. 1479 3.8609 -0.64
Farmer’s Club Member 0.8697 2.8896 0.30 1.7005 3.5090 0.48
Location-access characteristics
Distance to Bus Stop (km) 0.1355 0.2253 0.60 0.2401 0.3109 0.77
Distance to Market (km) -0.1 162 0.1923 -0.60 -0. 1946 0.2549 —0.76
Kasungu ADD 3.0240 2.3818 1.27 3.3523 2.5572 1.31
Blantyre ADD 7.1235* 3.7442 1.90 7.1028** 3.2291 2.20
Death Variables ( with Timin of years after a PA death)
PA male hos death -2.3915 3.0497 -0.78 —2.3980 3.4745 -0.69
PA female hos death -3.2161 3.5957 -0.89 -5.5741 4.8816 -1.14
PA male nhos death -0.7287 3.6233 -0.20 -0.8595 4.3814 -0.20
PA female nhos death -4.3177 2.8848 -1.50 -6.4583 4.61678 -1.40
Constants -- --
LR Chi Sq statistic 43.65 41.99
p-value 0.0081 0.0091
R-squared - 0.0195
Log likelihood -975.69 -1057.19
N 255 255

 

 

Source: MAPAMS 2002. *** 1% Significance, **5% significance, *10% significance
hos = female head or spouse, nhos = non head

Ho: The restricted model (Tobit) is better than double hurdle model

LLprobit = -458.04 , LLtrunc = -975.69, LLTobit = -1057.19

Lambda = -2(LLprobit + LLtrunc - LLTobit) = —2(-458.04-975.69 + 1057.19) = 717.08
Chi-square x2 0,, 22 = 33.92

Conc: Reject Ho. The double hurdle model is more suitable than the Tobit model

306

Table C3. Determinants of participation in rural non-farm employment by adult individuals in rural
households in Malawi, 2002 MAPAMS study. Model selection, Bivariate probit VS Probit

(with PA Death Variables)

Dependent variable: Participation (1=if individual participates; 0=otherwise)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Biprobit Regression
Rural Non-Farm Employment Rural Non-Farm Cassava Marginal
Employment Production effect

Independent Variables in 2002
Estimate Marginal Estimate Estimate
(robust std. effect (robust std. err.) (robust std.
err.) err.)

Individual Characteristics

Age .1173*** 0.0313 0.1160*** 0.0105 0.0003
(.0223) (.0201) (.0232)

Age squared -0.0013*** -0.0004 -0.0013*** 0.0002*** -3.09e-06
(.0003) (.0002) (.0001)

Education -0.0073 -0.0019 -0.0054 0.0480* 0.0003
(.0201) (.0194) (.0249)

Female -0. 1769 -0.0472 -0. 1756 0.0758 -0.00002
(.1451) (.1490) (.1694)

Household head 0.4195 0.1209 0.4666 1.2842* 0.0293
(.5959) (.5498) (.7706)

Household spouse -0.0432 -0.0114 -0.0317 0.2035 0.0016
(.2019) (.1954) (.2531)

Household Characteristics

Adults —0.1099*** -0.0293 -0.1 106*** -.0238 -0.0005
(.0328) (.0344) (.0389)

Dependency ratio -0.0520 -0.0139 -0.0525 -0.0081 -0.0002
(.0595) (.0598) (.0759)

Female headed 0.0824 0.0227 0.0719 0.4563 ,-0.0022
(.2337) (.2442) (.3613)

Age of head 0.0029 0.0008 0.0024 -0.0105 -0.0001
(.0118) (.0105) (.0134)

Education of head 0.0433 0.01 16 0.0395 -0.1079** -0.0007
(.0330) (.0321) (.0487)

Housing Value (MK) -2.628—06 -6.98e—07 -2.8e-06 -5.51e-06* -4.93e-08
(2.20e-06) (2.19e—06) (3.2e-06)

Land holding (ha) -0.0103 -0.0026 0.0014 0.1941*** 0.0014
(.0482) (.0494) (.0522)

Grow cassava in 2002 .2975“ 0.0876
(.1566)

Farmer’s Club -0.2740* -0.0662 -0.298* -0.7172*** -0.0034

membership (.1662) (.1554) (.2355)

Location-access characteristics

Distance to bus stop 0.0018 0.0005 0.0039 0.091 1*** 0.0007
(.0123) (.0132) (.0251)

Distance to market -0.0058 -0.0016 —0.0082 -0.0862*** -0.0006
(.0104) (.01 12) (0.02075)

Kasungu ADD 0.0915 0.0247 0.0620 -0.4589*** -0.0028
(.1153) (.1174) (.1379)

Blantyre ADD —0.3007** -0.0759 -2.4033*** -2.1 178*** -0.0137
(.1495) (.3488) (.4888)

 

 

 

307

 

Table C3. (Cont’d) Determinants of participation in rural non-farm employment by adult

individuals in rural households in Malawi, 2002 MAPAMS study. Model selection,
Bivariate probit VS Probit (with PA Death Variables)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Biprobit Regression
Rural Non-Farm Employment Rural Non-Farm Cassava Marginal
- Employment Production effect
in 2002

Estimate Marginal Estimate Estimate
(robust std. effect (robust std. err.) (robust std.
err.) err.)

Death variables

PA male hos death 0.3908“ 0.1193 0.3822“ -.1911 -0.0003
(.1735) (.1827) (.2983)

PA female hos death -0. 1536 -0.0384 -0.1473 0.1435 0.0006
(.2193) (.2257) (.3343)

PA male nhos death 0.1990 0.0572 0.1835 -5.889 -0.0062
(.2039) (.2133) (21 143)

PA female nhos death 0.3087* 0.0921 0.2961 -5.7509 -0.0059
(.1899) (.2266) (15699)

Constant -2.4569*** --- -2.4034*** -1 .245** ---
(.4524) .4254) (.4998)

Wald Chi Sq Statistic 225.35 271.27

p-value 0.0000 0.0000

R-squared 0.2239

Log likelihood -457.09 ~729.52

N 1119 1119

 

Source: MAPAMS 2002. *** 1% Significance, **5% significance, *10% significance

hos = female head or spouse,

nhos = non head

Summary test results for comparing probit vs bi-probit:

Artrho = 1596

Rho =0.1582 (.0876)
Ho: rho = 0

Chi-sq =3.1576

p — value =0.0756

308

 

APPENDIX D

Appendix D: Death Timing Model Results: rpam, epam and YSD

309

Table D1.

Determinants of participation in rural non-farm employment by adult individuals in

rural households in Malawi, 2002 MAPAMS study. Probit Estimation (with rpam, epam
and YSD Variables)

Dependent variable: Participation (1=if individual participates; 0=otherwise)

 

 

 

 

 

 

 

Rural Non-Farm Agricultural Non-Agricultural
Employment RNFE RNFE

Independent Variables Estimate Marginal Estimate Marginal Estimate Marginal
(robust std. effect (robust effect (robust std. effect
err.) std. err.) err.)

Death Timing

rpam 0.7 88** 0.268 0.043 0.005 0.700** 0.170
(.339) (.375) (0.328)

epam 0.323 0.098 -1.122 -0.061 1.074 0.303
(.287) (1.209) (.870)

Years since death -0.001 -0.0002 0.105 0.013 -0.066 -0.011
(.103) (0.133) (0.102)

 

 

 

 

 

 

 

Source: MAPAMS 2002. *** 1% Significance, **5% significance, *10% significance

Result Table D2. Determinants of intensity of participation in rural non-farm employment by adult
individuals in rural households in Malawi, 2002 MAPAMS study. Truncated regression

(with rpam, epam and YSD Variables)

Dependent variable: Off-farm work time (number of hrs worked per week)

 

 

 

 

 

 

 

Rural Non-Farm Agricultural Non-Agricultural

Independent Variables Employment RNFE RNFE
Marginal Marginal Marginal
Effect Z Effect Z Effect Z
(robust std. Statistic (robust std. Statistic (robust std. Statistic
err.) err.) err.)

Death Timing

rpam -3.960 -1.23 -14.49* -1.76 -5.685 -1.39
(3.208) (8.238) (4.097)

epam 17.164 0.68 -23.547 — l .39 37.385 0.88
(25.172) (16.89) (42.71)

Years since death —0.702 -0.56 3.182 1.51 -1.655 -0.90
(5.63) (2.111) (1.845)

 

 

 

 

 

 

 

Source: MAPAMS 2002. *** 1% Significance, **5% significance, *10% significance

310

 

 

 

 

 

UBRARIE.
Ii‘

YATE UNIVERSITY

5

N
mu.
H
Flu
M