I'll
Hi Ni! '1

1"“14 H‘

T

I

W

\‘T
M
l
‘T
T“
T
‘l
Ti

IT

IT

 

_ THE IMPACT 0F ADULT DEATH
on. RURAL SMALLHOLDERS IN KENYA: .

IMHICATIONSFORMWAIDS '

[PianB Paper for the flagree 055:1; 8.. ' ‘

. MICHIGAN STATE UNTVERSTTT ' ‘

TTELTJTTY TessaETAT-TMC HER . _ ~

2002 '.

 

 

LIBRARY
Michigan State
University

 

 

 

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

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/DateDuepss-sz

 

THE HVIPACT OF ADULT DEATH ON RURAL SMALLHOLDERS IN KENYA:
IMPLICATIONS FOR HIV/AIDS

By

Melody Rebekah Mc Neil

A PLAN B PAPER

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE
Department of Agricultural Economics

2001

ABSTRACT

THE IMPACT OF ADULT DEATH ON RURAL SMALLHOLDERS IN KENYA:
HVIPLICATIONS FOR HIV/AIDS

By
Melody Rebekah Mc Neil

The incidence of prime-age adult death is increasing as a result of HIV /AIDS and
opportunistic infections. Because of these increases, the ability of smallholders to
increase food security and maintain current on and off-farm activities is more constrained
than ever. Using panel data, this paper empirically examines the effects of disease-related
adult death on small farm households in Kenya. With increasing pressure on land and
resources, the strategies employed by these households are currently centered around
maintaining household composition consistent with their needs for household and farm
labor. The demographic crisis being faced in some areas of Kenya is not repeated across
all districts at this time. Government and non-govemment entities must shape policy and
programs to stop adult death from becoming an extensive crisis, and to off-set the food

insecurity created by loss of productive labor and increased dependency ratios.

ACKNOWLEDGMENTS
I would like to acknowledge my major professor, Thomas Jayne, and my
committee members, John Strauss, David Tschirley, and Takashi Yamano for their
guidance. Thank you also to my fellow students at MSU for their input, especially Amy
Damon, Corey Risch, Eric Knepper, and Gerald Nyambane. Finally I want to express
love and gratitude to Joseph, Diane and Geoffi'ey Mc Neil, Jennifer and Michael Johnson,

and to Keith for loving support, listening ears, and for believing in me.

TABLE OF CONTENTS

List of Tables ...................................................................................................................... iii
List of Figures ..................................................................................................................... iv
1.0 Introduction .................................................................................................................... 1
2.0 Literature Review ........................................................................................................... 5
2.1 Household Level Effects .................................................................................... 6
2.2 On-Farm Effects ................................................................................................. 9
2.3 Off-Farm Effects .............................................................................................. 11
3.0 Data .............................................................................................................................. 14
3.1 Inflation Discussion ......................................................................................... 15
3.2 Construction of Variables ................................................................................ 15
3.3 Characteristics of Sampled Households ........................................................... 19
4.0 Analytical Framework .................................................................................................. 31
4.1 Household Model ............................................................................................. 31
4.2 Household Framework ..................................................................................... 34
5.0 Empirical Specification ................................................................................................ 42
5.1 Description of Variables .................................................................................. 43
6.0 Results .......................................................................................................................... 50
6.1 Household Demographics ................................................................................ 51
6.2 Off-Farm Effects .............................................................................................. 57
6.3 On-F arm Effects ............................................................................................... 60
7.0 Conclusions .................................................................................................................. 63
Appendices ......................................................................................................................... 67
Bibliography ....................................................................................................................... 86

ii

LIST OF TABLES

Table 1. % of Households Suffering an Adult Death by Province and Cause ................... 20
Table 2. Income Characteristics of Deceased Household Members .................................. 21
Table 3. Household Characteristics of Sampled Households ............................................ 21
Table 4. Comparative Demographics of Adult Household Members, 2000 Survey ......... 23
Table 5. Demographic Characteristics of Sampled Households by Province .................... 23
Table 6. Demographic Changes from 1997 to 2000 Number of Individuals Fitting the
Category ............................................................................................................................. 24
Table 7. Changes in Farm Characteristics Between 1997 and 2000 .................................. 26
Table 8. Changes in Off-farm Activities and Income from 1997 to 2000 ......................... 27
Table 9. Dependency Ratios .............................................................................................. 28
Table 10. Characteristics of New Members by Gender 2000 ............................................ 29
Table 11. Characteristics of New Members Reporting Non-farm Income in 2000 by
Gender ................................................................................................................................ 29
Table 12. Characteristics of New Members ....................................................................... 30
Table 13. Distribution of Deaths by Year of Death ........................................................... 34
Table 14. Definition of Variables and Descriptive Statistics ............................................. 48
Table 15. Predicted Effects of Adult Deaths on Household Composition ......................... 52
Table 16. Predicted Effects of Adult Death on Off-farm Household Indicators ................ 57
Table 17. Predicted Effects of Adult Death on Farm Household Level Indicators ............ 62

iii

LIST OF FIGURES

Figure 1. Timing of Illness and Death ............................................................................... 33
Figure 2. # of Children Living in a Household .................................................................. 34
Figure 3. Caregiving Needs ................................................................................................ 35
Figure 4. Sum of Hired and Family Labor ......................................................................... 36
Figure 5. Value of Assets ................................................................................................... 38
Figure 6. Off-farm Income ................................................................................................. 39
Figure 7. Farm Income ....................................................................................................... 40

iv

1.0 INTRODUCTION

Rural people in Sub-Saharan Africa are in a never-ending cycle of balancing the
use of their limited resources to best fit their consumption, health, and social needs.
Unforeseen death and other forms of severe household stress, catalyze people to trade-off
between conflicting needs in order to continue to engage in the activities that best support
them. Researchers have only a limited empirical understanding of how rural households
cope with the adverse shocks to their livelihoods. Coping can be defined as “an array of
short-term strategies adopted in response to crisis.” Responses to crises can be
understood as attempts to minimize not only the immediate effects of crises but also their
long-term effects on the household (Adams, Cekan, and Sauerborn, 1998).

Food security, defined as access to enough food by a household for a healthy and
active life, is a major concern of smallholders. With the advent of household level crisis,
households will face the need to redistribute labor and responsibilities in order to try to
maintain their welfare subject to a potentially more restricted opportunity set. The death
of a family member is one important type of a household level crisis. Unexpected adult
death has become a more salient type of crisis in Africa with the rise of the HIV/AIDS
pandemic. Understanding the potential impacts of adult death allows policy makers as
well as other groups to shape their attitudes and actions to more appropriately match what
households will go through as the crisis becomes larger.

Adult death will become more common as prime-aged adults become more
heavily affected by AIDS (World Bank, 1997). The growth of an AIDS crisis will also

make other diseases such as tuberculosis more common and uncontrollable. Any

household-level effects of adult death will have important implications for the expected
spread of AIDS. HIV/AIDS is a global epidemic affecting people in every part of the
world. While AIDS is not necessarily a disease of poverty, it appears to have grown most
rapidly in relatively poor parts of the world. Since poorer households suffer more intense
and persistent effects of adult death from HIV /AIDS and other causes, all productive
sectors have a role to play in combating AIDS, including agriculture, non-farm sectoral
growth, health and education (Stillwaggon, 2000, World Bank, 1997, and NASCOP,
1999)

F AO estimates that about one tenth of world population lives is Afiica; however,
about 83% of AIDS related deaths, as of 2000, had occurred in Afiica (FAO, 2000).
AIDS’s impact on farming communities and systems differs widely from place to place.
Nevertheless, it is apparent that HIV /AIDS is and will continue to constrain the
developmental process in many LDC’s, specifically in Sub-Saharan Afiica (SSA).
Positive trends previously achieved in child health seem to be slowing and perhaps even
reversing with the growing AIDS epidemic (Ainsworth and Semali, 2000).
Understanding the potential household level shocks brought on by HIV/AIDS is crucial in
order to develop and implement strategies and extension services that effectively deal
with the AIDS epidemic in economic and agricultural development.

Economic costs may paralyze a household as an ill member loses income, their
medical expenses increase, education of other members is foregone in order to continue
agricultural activities and care for the sick. Both household production constraints and

time constraints may be altered given the loss of resources associated with a particular

household member. Resource and land endowments could be permanently altered with
the death of an adult, especially if the adult is a household head. Human capital
possessed by the deceased may change production constraints, as the technical know how
or management capacity of a household changes (Bollinger and Stover, 1999).

Using panel data collected from households in rural Kenya in 1997 and 2000, this
paper specifically examines how adult deaths, which are increasing as a result of AIDS,
affects smallholder household activities. We determine the effect of a death of a
household member on income and asset values as well as entry of new members, or the
return of former resident members, into a household. We estimate this relationship while
controlling for characteristics of the deceased person, such as age, gender, level of
education, and interactions between these attributes. The household fixed-effects
approach provides information on differences in household behavior and welfare between
a period before adult death with a period afier the occurrence of adult death, and helps to
control for unobserved time-invariant household characteristics. Increased analysis of
household behavior before and after adult death is important to predict the potential
impacts of HIV/AIDS in a variety of settings.

In this paper we examine the effects of adult death from disease in general rather
than focusing our attention on AIDS-related adult death. A paucity of information in the
household surveys preclude differentiation between AIDS-related deaths and those caused

by other diseases. Ainsworth and Semali (2000) examined the effects of adult death in

general for the same reason.1 While adult death and AIDS-related adult death are not
explicitly separated in this analysis, many conclusions can be drawn from this study
about the possible effects of AIDS-related deaths on household behavior. However, it is
aclqrowledged that the cause of death may have some differential effects on household
behavioral responses and welfare. Furthermore, given the problem of unexpected or
sudden adult death, an empirically-based description of rural household responses to a

death in the family is important for the design of future rural development strategies.

 

' The survey data used in this paper does ask respondents to identify any deaths incurred in the
household by accident, disease, or old age, but such information is typically regarded as unreliable.

4

2.0 LITERATURE REVIEW2

Many factors affect the behavior of smallholders in the face of adversity.
However, there is limited quantitative information available about actual household
response and the factors that cause differences in responses across households. Adult
death affects household behavior at three distinct levels classified as the attributes of (1)
the deceased individual; (2) the household; and (3) the community. Not only does the
impact of death on the remaining household members differ by level, but coping
strategies of the household unit will also vary depending on perspective (World Bank,
1997). At the individual level, coping may consist of doing less work at home to
maintain a certain standard of productivity on the job. At the household or community
level, those who fill in for an ill or injured adult may be drawn away from other
productive activities. Investment in human and physical capital may also be sacrificed
(Ainsworth et a1, 2000, McPherson et al, 2000, and Over et a1, 1992).

According to one source the percentage of adults in SSA who are presently HIV
positive is 8.57% (Committee on World Food Security, 2001). In Kenya the percentage
of adults (ages 15-49) who are HIV positive is reported to be 14.0%, much higher than
the regional average (UNAIDS, 2000). While development agencies may be generously
aiding in campaigns to educate and cope with the AIDS epidemic, they are not

concurrently scaling down their other development projects to mirror this expected loss of

 

2 This section, and following sections, include both empirical and anthropological information.
Empirical results from Kagera remain the most important work done in quantifying the effects of
HIV/AIDS to date. Published results of the Kagera study do have details which allow analysis of
household coping with an AIDS death. Although data are still quite limited, the authors expect their
findings to be consistent with further studies (World Bank, 1997).

5

adult life (Bonnel, 2000). However, a consequence of increased mortality of prime-aged
adults in the next 10-20 years is higher dependency ratios of children and elderly people
(McPherson et a1, 2000).

The expectations for changes in income and income generating activities due to the
incidence of adult death, changing demographics, and resource levels are explored
empirically in the following sections of this paper. Hypotheses fall under three general
categories: 1) household demographics, 2) on-farrn effects, and 3) off-farm effects, and as
such the literature review is structured similarly.

2.1 Household Level Effects

Farm household performance is dependent on the characteristics of each member.
The loss of one has ripple effects for the entire household. A death will reduce a
household’s stock of human capital as well as time and disposable resources (du Guemy,
2001). The need to alter spending and labor allocation patterns to consider the needs of
ill and ailing adult members in light of other household responsibilities, strains more than
just one household. However, the ultimate death of an adult will also release the labor
that had been taken up in caring for that individual (White and Robinson, 1999).

Illness also generates the need for increased expenditure on already stretched medical
resources. Kenya has spent as much as 74% of a yearly health care budget on hospital
care in more than one fiscal year over the past two decades (Over et al, 1992). An
intensified AIDS crisis will only worsen the disparity in health spending, pulling it further
away from primary care. Also, a widespread epidemic may jeopardize local and national

govemments’ ability to continue adequately providing social services (Nampanya-Serpell,

2000, Bonnel, 2000, Committee on World Food Security, 2001, University of Nairobi,
1997, World Bank, 1997).

Household health may suffer even more, as labor shifts away from hygiene
maintenance chores such as cleaning and proper food preparation towards caring for the
ill (Over et a1, 1992). This is expected to have an especially negative impact on child
nutrition and health. For instance, stunting is more common in children whose parents
have died, and evidence shows the death of a father can be equally as detrimental as the
death of a mother. The death of a female adult has also been shown to increase the
likelihood of older children discontinuing education and to raise infant mortality
(Ainsworth, Beegle, Koda, 2000, Ainsworth and Semali 2000, Lundberg and Over,
2000)

While children are a main concern in this issue, women tend to be more vulnerable to
the effects of disease in their households than men. Women typically carry out the task of
tending to the ill as well as other dependents. They also continue carrying heavy
workloads beyond these extra demands (Committee on World Food Security, 2001).
Ironically desired family size increases as the need to share work also increases, but the
potential health of those family members decreases (Over et a1, 1992).

Changes in household composition can help to ameliorate the effects of an adult death
on households. In Kagera, Tanzania households suffering an adult death showed very
slight changes in average household size, from 6.0 to 5.7 members. A similar result was
shown in Rakai, Uganda. Child dependency ratios, defined as the number of children in

a household divided by the number of adults, were found to increase only very slightly in

Kagera, from 1.2 to 1.4, compared to from 1.2 to 1.5 for households not suffering an
adult death (World Bank, 1997). Furthermore, recovering households were more likely to
be headed by women than those households who had not suffered an adult death in
Kagera, 36% compared to 25% (World Bank, 1997). This result has serious implications
for the health and well-being of other household members as it has already been stated
that women are generally poorer than men. Moreover, if the head of household died
from a communicable disease, the surviving spouse is likely to face health risks as well.

Responding households in Chiang Mai, Thailand showed different results. In
households suffering a death in Chiang Mai, average household size was reduced by one
full person, even afier two years (World Bank,1997). This result could stem from higher
incomes in Chiang Mai than the surveyed areas in Afiica. It also suggests different
patterns of coping with household crisis. Household composition is expected to play a
major role in the coping efforts of households trying to cope with an adult death in SSA.

As the crisis of adult death continues to grow in the Sub-Saharan region, the trends
that have been observed in other reports concerning the household level responses to
adult death may begin to show signs of breaking down. The likelihood for the average
number of adult deaths per household to increase over the course of 2-3 years,
considering the impact of HIV/AIDS and its accompanying opportunistic infections,
should decrease. Continuing to give equal weight to AIDS related and other disease
related deaths will be important in monitoring the household composition patterns of
those coping with adult death, as secondary infections of HIV/AIDS become more

widespread and malaria continues to take lives.

2.2 On-Farm Effects

On-farm effects are observable in some recent studies and are expected to become
more pronounced within the next 1-2 decades. It is estimated that of the 14 million
AIDS victims in SSA, at least 60% were engaged actively in agriculture. To lose no less
than 25% of SSA’s agricultural labor force over the next 20 years would not be unlikely
(Development News, 2001).

The factors leading to farm-level changes as a result of an intense adult death crisis
are varied. One community level constraint to agriculture in Kagabiro village, Tanzania,
is compulsory funeral attendance. More importantly is the moratorium on all farming
activities for 2-3 days and on planting anything for the 7 days of the mourning tradition.
Researchers in that area say that the time dedicated to mourning deaths in the community
can amount to 4-8% of yearly labor time (Tibaijuka, 1997). Many of the respondents to
the study done in Tanzania indicated that they were concerned over their food security
leading up to and following an adult death as a result of women planting fewer root cr0ps
(T ibaijuka, 1997). This is attributed to lack of labor to cover other household activities.

Substitution of labor-saving subsistence crops for labor intensive crops has been
identified as a strategy that many households employ to deal with the hardship of an adult
death (Committee on World Food Security, 2001). Labor intensive crops may seem less
feasible in light of adult death (UNAIDS, 1999). In Rakai District in Uganda, there is
already strong evidence of a switch to subsistence types of crops, likely resulting from
increasing rates of adult death. Whereas in 1989 there was a great deal of savory banana

cultivation accompanied by many different types of relishes, in 1993 there were far fewer

relish crops planted (Bollinger, Stover, and Nalo, 2000).

Increased subsistence production was more prevalent in more impoverished
households. Small farms are typically less able to absorb the shocks to adult labor
availability associated with adult death. Farm families are also shifting away fi'om food
crops high in protein and other nutrients to root crops that are easier to produce
(Development News, 2001 , du Guemy, 2001). They have fewer resources to reallocate
and opportunities to hire labor and animal power, and are rather selling off livestock
including draft animals. (Development News, 2001, and Bollinger, Stover, and Nalo,
2000)

A shrinking labor base may also lead to abandoning cultivation of more remote fields.
As some fields lay fallow because there is not enough time or manpower to cultivate
them, food security and production may be further compromised (du Guemy, 2001). An
alarming trend is less emphasis on conservation of land. In order to maintain survival,
concern over long-term sustainability of farm land may decrease. Commercial production
is also expected to suffer as the loss of seasonal workers may intensify and morbidity and
mortality of employees may increase (Committee on World Food Security, 2001 and
du Guemy, 2001). Some of these coping tactics are those usually expected only in famine
periods, and are associated with a higher opportunity cost of adult labor as a result of
AIDS.

Other strategies have also had an important impact on households. Dissaving and/or
selling assets are also widespread (du Guemy, 2001, Tibaijuka, 1997, and World Bank,

1997). Studies in Kagera, Tanzania, Rakai Uganda and Chiang Mai, Thailand show

10

dissaving. In Kagera and Rakai selling durable goods like radios was very common in
households recovering from an adult death. The Kagera study reported recovering
households membership in credit groups decreased more than non-recovering households
(51-36% compared to 41-36%). Chiang Mai results showed land sales were also a
strategy (World Bank, 1997).

Although empirical results are shown from three studies, Kageara, Chiang Mai and
Rakai, it has already been stated that data continue to be limited. Chiang Mai results do
not have the ability to compare financial variables between households experiencing and
those not experiencing death (World Bank, 1997), such that the characteristic
comparative nature of this analysis remains an important exercise. Furthermore, much of
the published material on this subject remains focused on impacts of adult death on
children and education rather than cross-cutting farm-level issues.

2.3 Off-Farm Effects

There are many potential pathways by which adult death may affect households’ labor
productivity, although solid empirical information is so far lacking. Growth in
production and labor are expected to decline along with the shifting of production
functions. There is also an expected decline in human capital stock as well as lost
incentives to continue investing in human capital (du Guemy, 2001, and Bonnel, 2000).
The continued education of children and opportunities of transferring cultural awareness,
knowledge of common practice and traditions is declining.

Through village case study, Adams (1998) finds that in Malawi’s rainy season, death

of a productive adult worker can limit a household’s productivity and income

11

opportunities. Both food security and future productivity are priorities for household
strategies dealing with adult death. In his case study, Mtika (2000) shows that during the
rainy season families may send a worker off-farm in the event of an adult death. Sending
a member away from the household will not necessarily eliminate his or her claim on
household resources; however, it does imply a possibility for some remittances and
perhaps a higher return to that person’s labor (Adams, Cekan, and Sauerborn, 1998).

Help from neighbors and family may be forthcoming. According to the study done in
the Kagera region of Tanzania, households in crisis were likely to receive in-kind or cash
assistance from other households (World Bank, 1997). This study and others show that
while a household suffers great loss after the death of an adult member, they also suffer
during the illness, and transfers from other households tend to be greater after the death of
a household member not during his or her illness (World Bank, 1997). The average
length that an AIDS patient, for example, would spend bed-ridden in Kagabiro village,
Tanzania is 6 months (Tibaijuka, 1997). This presents an even greater problem for the
suffering household as a projected average of 29% of family labor stock was dedicated to
the care of an AIDS patient (Tibaijuka, 1997).

A quantitative study on adult death in Cote d’Ivoire shows that in the ten months after
the death of an adult member, household consumption dips and then partially recovers.
Similar patterns are seen in studies done in Rakai, Uganda, Chiang Mai, Thailand,
although these studies cannot confirm the results (World Bank, 1997). Also, the Kagera
study shows that households recovering from an adult death consume a smaller amount of

purchased food compared to households who had not incurred an adult death in the same

12

period. The authors found this supported the view that recovering households worked a
lower number of hours for wages, and therefore had lower incomes (World Bank, 1997).
The study from Kagabiro village, Tanzania also indicates a distinct increase in the cost of
hired labor over the study period (1982-90), while farm incomes have decreased in real
terms at the same time.

The impact of adult death on off-farm productivity cannot be underestimated as the
studies discussed above have indicated that preservation of food security is the first
priority and off farm activity will decrease for that purpose. Comparison of income
variables over the two sets of households presented in this paper will show quantified

results relating to these concerns.

13

3.0 DATA
This section presents the data set used in the empirical analysis. Quantitative analysis

will be carried out using data from two linked household surveys conducted by Tegemeo

Institute of Egerton University and Michigan State University. Both surveys visit the

same sampled households. The first was completed in September 1997 and the second in

July 2000; the recall period of both surveys covers the same crop season and marketing

year. The surveys are designed to be representative of the main agricultural regions of

the country (Tegemeo Institute, 2000).

The sample size of the 1997 survey was 1,574. In 2000, 1,443 of these households
were revisited fi'om the original 1997 sample. The attrition rate, therefore, is 6.3 percent.
In the 2000 survey, 31 additional households were added as replacements. But, for
comparison, we do not include them in the following analysis. The sample size was
further reduced because of the following:

A. Two primarily pastoral districts included in 1997 (Garrisa and Turkana) were
excluded in the 2000 survey. The numbers of original households excluded are 40 in
Garrisa and 34 in Turkana.

B. Thirty-one additional households were excluded because of incomplete or unreliable
information. This reduced the sample to 1,412 households interviewed in both 1997
and 2000.3
Information collected in the survey lends itself to the comparison of household-level

changes in a before death/after death scenario. After controlling for other household-

 

’ I thank Takashi Yamano for his help in cleaning the data.

14

level and community-level factors through econometric analysis, the “before/after”
analysis can be strengthened to a “with/without” analysis. Household-level panel data
also controls for the effects of time-invariant unobserved variables, thereby mitigating the
common omitted variable problem of cross-sectional analysis. Detailed income
information was collected in both surveys, such that disaggregation up to crop income,
non-farm income, remittances, and livestock income are comparable.
3.1 Inflation Discussion

In order to decrease the possibility of any bias caused by inflation between the two
surveys, all 1997 price and income data was inflated to match the 2000 data. The inflator
was calculated using monthly CPI data from Central Bureau of Statistics’ Economic
Digest 2000 (Government of Kenya 2000). Averaging the CPI from the 12-month recall
period recorded in the two surveys was the method used to calculate the actual inflator.
The recall period for the 1997 survey was September 1996-August 1997. The recall
period for the 2000 survey was August 1999 to July 2000. The average of the indeces
from the period covering the survey was 1.177. All price and income variables fiom
1997 were inflated using this number to derive 2000 constant prices.
3.2 Construction of Variables

Two kinds of variables are used in the econometric model presented in section 5.
Both variables in levels and differences are used in the estimations.
3.2.1 Variables in levels and categorical variables

All variables presented in levels are taken from the 1997 data set. The variables

include the characteristics of the deceased individual such as that person’s age or

15

educational level at the time. These variables are taken fi'om individual information in
1997 and are recorded as continuous variables. Educational level is referred to as number
of years of education. Age is a continuous variable indicating the actual age of the
deceased person in 1997. There are also other crucial variables included as categorical
variables. The most notable of these is the adult death variable itself. This variable was
calculated as taking a value of 1 when a disease-related death has occurred in a household
and 0 when a disease-related death has not occurred. Disease-related death accounted for
92% of the total adult deaths in the sample.

An important distinction to be made concerning the adult death variable is that all
reported disease-related deaths are included in this analysis regardless of how many
months the deceased individual may have spent on farm in the previous 12 months. The
rule of household membership used for all other individuals is those individuals living on
farm for 5 or more months. This did not apply to deceased individuals for several
reasons. Most importantly is their demonstrated importance to the household for income
as shown in table 2. However, a more important reason for their inclusion regardless of
time on farm is the inability of the researcher to account for the actual date of their death.

With other household members it is possible to tell between 1997 and 2000 if they
have actually increased their number of months on farm. However, if a deceased member
spent less than 5 months on farm in the 1997 enumeration period, it is impossible to tell if
this pattern changed in the following three years if they died before 2000. It is also not
possible to tell which month of the first enumeration period 1997 deaths occurred in.

Furthermore, a striking characteristic of the 19 deceased individuals from the 2000 survey

16

is 1 is reported to have been on farm for less than 6 months. This indicates that an ill
member may come home while ill to be cared for, effecting household decisions during
that time. For these reasons all reported deceased members are included in this analysis.

Categorical variables included in the models are year of death variables, relationship
of deceased member to the household head, and gender of the household head.
3.2.2 Diflerenced Variables

The construction of differenced variables is straight-forward: 1997 values were
subtracted from 2000 values of household information included in the specifications.
However, the construction of the components of the differenced variables are important to
note.
1. Household income
Because of differences in the collection and recording of data, it was important to
construct comparable measures of income between the two periods. Important notes to
be made on the construction of a comparable data set for the two surveys are:
a. The 1997 survey did not ask for the cost of hired farm labor. Therefore the 2000
survey data did not subtract out hired labor costs in the calculation of farm income. Other
cash input costs and land preparation costs were deducted. Livestock income is gross
income, the sum of the sale of animals and the sale of livestock products.
b. The categories of off-farm or informal income were not the same between the two '
surveys, such that there could not be a high level of disaggregation of income by source.
Aggregation was done by category of off-farm work. The aggregation of categories made

it possible to derive comparable measures of farm labor wages and business income.

17

Business income includes income from self-employment in off-farm activities.
Furthermore, detailed cost information was not collected in 1997; so, off-farm income is
compared between the two years in gross rather than net terms. This ensures the
comparability of the two data sets. Final off-farm income is calculated as the sum of
informal income, salary, remittance and pension income.

c. Constant prices are used across the two periods to value crop income. If the objective
is to understand how adult mortality affects household behavior, one might choose to
hold prices constant across the two survey periods, forcing changes in crop income to
reflect changes in quantities harvested. This is because crop prices are exogenous from
the standpoint of the household; households only have influence over physical production
decisions, and even then only partially. For the main analysis in Chapter 5, I use 1997
district-level crop prices to value household crop production. However, to assess the
robustness of the findings, I present the income measures using 1997 prices in 1997 and
2000 prices in 2000.

d. Information concerning fodder was well collected in 2000, but not 1997. Therefore,
fodder has been dropped from both crop income measures.

e. Total household income is used in the analysis; however, it is important to stress that
farm income is net of cash input costs, while non-farm income is measured in gross
terms. This weights the total income variable disproportionately toward non-farm
activities.

f. Finally, it is important to note that remittances were collected in 1997 as a combination

of remittance, pension and rrriscellaneous other income. To ensure comparability of 1997

18

and 2000 rerrrittance income measures, pension income was added into the remittance
income measure for 2000.
2. Demographic Characteristics
Several household level demographic characteristics are differenced for the estimations
reported in section 6. The methods for calculating these variables are discussed here.
Please note that the deceased household member was not included in the differencing
process for econometric estimation. The interest is in the movements of other household
members in and out of the household. Therefore, the 1997 value of household size (hhsz)
is calculated as hhsz-l when a household has a death recorded and hhsz when a
household doesn’t have a death recorded. Similarly, when this variable was
disaggregated to adult females, males and number of children, 1 was subtracted fiom the
1997 observations when a death of that gender was recorded in a household. This same
method was followed when constructing tables.
3.3 Characteristics of Sampled Households

This section presents a discussion of some important characteristics of the
interviewed households. Each of the tables highlights specific points about the data that
will make evident the disparity between households experiencing adult mortality and
those that do not. It also gives insight into the adjustment pattern that households may
undertake in order to maintain their integrity during and after a crisis.

Table 1 shows that of 1412 households included in the total sample the most
important cause of adult death is disease. The provinces suffering the highest incidence

of adult death are Nyanza and Coast. Siaya and Kisumu, both districts in Nyanza, are

19

believed to have some of the highest rates of HIV/AIDS prevalence in Kenya. However,
both are also well known malaria areas. In Nyanza province household size is below the
national average while the number of children in the household remains close to or above
the national average. This does give support to current literature that says households are
likely to increase their number of children in order to increase labor availability.
Fostering children to make-up some extra labor is also an important possibility resulting
from widespread disease related adult death.

Table 2 highlights the income of the deceased members as recorded in 1997. The
table indicates that the deceased members were still contributing a substantial amount to
household income. Of the 89 deaths reported in this paper, 21 had income in 1997. Of
these 21, 13 were from either Nyanza or Coastal provinces. These are the provinces with
the most adult death fi’om disease in this survey.

Table l. % of Households Suffering an Adult Death by Province and Cause

 

Cause of Death Total Number of Total Number of
Province (% of Households) Total Deaths Disease related Deaths
Disease Other (%)

Coast 7.14 1.14 8.33 7 6

Eastern 3.90 ' 0.43 4.31 10 9

Nyanza 14.50 0.76 15.27 40 38

Western 5.21 0.35 5.56 16 15

Central 3.45 0.57 4.02 7 6

Rift Valley 4.03 0.53 4.57 17 15

Total 6.30 0.56 6.87 97 89

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegemeo Institute, Nairobi, Kenya.

20

Table 2. Income Characteristics of Deceased Household Members (Off-farm Income
measures are averaged for only those deceased members reporting that category of

income in 1997)

 

% of 1997 % of 1997 farm % of 1997 % of 1997
members informal members wage members salary members remittances
with off-farm with income with (mean with (mean Ksh)
zero off- income zero 06- (Ksh) zero Ksh) zero
farm (Ksh) farm values values
income income
values
Male (n=46) 86.96 73,090 97.83 5,000 82.61 70,880 93.48 1 1,800
Female 90.70 24,023 0.00 0.00 97.83 21,600 90.70 1 1,400
(n=43)
All (n=89) 88.76 53,463 98.90 5,000 88.90 65,404 92.13 11,571

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegemeo Institute, Nairobi, Kenya.

Table 3. Household Characteristics of Sampled Households (2000) by Province (all
numbers are household averages)

 

 

 

Coast Eastern Nyanza Western Central Rift Valley National
Number of Household Members
Children
Below 6 0.81 0.78 0.82 1.18 0.42 0.85 0.84
Girls 6-16 1.27 0.83 0.96 1.20 0.70 1.09 1.06
Boys 6-16 1.36 0.81 1.04 1.32 0.66 1.13 1.01
Adults
Women 2.27 1.82 1.67 2.08 1.73 1.92 1.89
Men 2.11 1.53 1.52 1.92 1.68 1.88 1.75
Houshold Size 8.36 5.74 6.03 7.53 5.20 6.85 6.54
Total Acres Cultivated 5.54 6.36 3.77 5.37 3.21 6.02 5.15
Value of Assets (Ksh) 38,179 64,568 41,089 56,295 58,760 16,2905 84,256
Members engaged in off- 2.80 1.92 1.23 1.49 1.33 1.53 1.58
farm Employment

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegemeo Institute, Nairobi, Kenya.

21

Table 4 shows that all adults in households incurring and not incurring adult deaths
had spent most of 2000 within the household. This is a strong indication that household
members have been identified as those who actually inhabit the household a majority of
the time. These members are, on average, over 30, and the women have generally less
education than the men for both household incurring (6.26 years for women, 7.31 for
men) and not incurring (6.28 years for women, 7.45 for men) adult death.

For many households that incurred a death in the family since the 1997 survey, there
were probably strong effects from any long illnesses effecting the 1997 data. The 1997
data did not collect information on illness in the family. Hence, the effects of illness may
already be picked up in the 1997 for some households that later incurred a death. Again
the figures in Table 5 do not show large shifts, however it is impossible to characterize
the exact movement of the demographic changes from 1997 to 2000 without knowledge
of exact timing of illness and death. Table 5 shows that across all provinces, the number
of adult household members increases fiom 1997 to 2000, but that the number of children
decreases. Then table 6 further supports this at the national level. Even in households
recovering from an adult death, the number of men and women members only decreases

slightly between the two surveys.

22

Table 4. Comparative Demographics of Adult Household Members, 2000 survey

 

Average Relationship to household head Average # months
age Head Spouse Son/ Other Others years of spent on
Daughter Relatives education farm in
‘34. last year

 

 

Households lncuning an Adult Death

Female 34.13 10.43 38.04 1.23 33.13 20.25 6.26 11.66
Male 32.13 24.24 1.21 O 56.97 18.79 7.31 11.33
All adults 33.14 17.12 19.22 0.60 44.44 18.62 6.78 11.50

Households not Incurring an Adult Death

Female 34.85 6.94 42.33 3.11 36.15 11.47 6.28 11.64
Male 34.60 37.67 0.75 0.67 45.90 14.76 7.45 l 1.49
A11 adults 34.73 21.90 22.10 1.92 41.40 13.07 6.85 l 1.57

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegemeo Institute, Nairobi, Kenya.

Table 5. Demographic Characteristics of Sampled Households by Province

 

Province Coast Eastern Nyanza Western Central Rift Valley

 

Percent of 6% 4% 14% 5% 3% 4%
Households

incurring an

adult death

 

Average # 2.11 1.38 1.23 1.71 1.59 1.65
Men 1997

Average # 2.11 1.53 1.52 1.92 1.68 1.88
Men 2000

Average # 2.17 1.75 1.34 1.75 1.62 1.67
Women 1997

Average # 2.27 1.82 1.67 2.08 1.73 1.92
Women 2000

Average # 4.05 2.54 3.19 3.66 2.15 3.44
Children
1997

Average # 3.44 2.42 2.82 3.70 1.78 3.06
Children
2000

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegemeo Institute, Nairobi, Kenya.

23

Table 6. Demographic Changes from 1997 to 2000 Number of Individuals Fitting
the Category (differentiating between households experiencing adult death and those not
experiencing adult death and including the deceased member in 1997)

 

1997 2000 Average 1997 2000 Average
Average Average change Average Average change
Variable w/out death w/out death w/death w/death
Women 1.67 1.89 0.22 1.95 1.87 -0.08
Men 1.58 1.75 0.17 1.84 1.79 -0.05
Children 3.17 2.90 -0.27 3.12 2.96 -0. 15

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegemeo Institute, Nairobi, Kenya.

Table 7 begins a detailed look into the changes in income generating activities
undertaken on and off the farm by surveyed households. It is important to stress here that ‘
households who have recorded an adult death may have been affected in many ways
before the actual event of a death. Asset values alone indicate a large difference between
the 1997 values for households with and without an adult death. Dissavings and
disinvestrnent are both expected for households with a terminally ill individual.
Therefore, the 1997 values of some of the variables in Table 7 may already reflect to
some extent the effects of illness, at least for the subset of households in which the cause
of death was a disease that affected the household’s behavior even before the 1997
survey.

Farm wage labor income increases for families who have suffered an adult death
while it shows a decrease for those families not incurring an adult death. Again this
supports the assumption that a terminally ill household member induces a household to
make decisions concerning labor allocation, savings, and investment contrary to the
decisions they would make under normal circumstances. The increase in farm wage labor

income indicates a greater number household labor hours being dedicated to wage labor.

24

The extra hours may come from a household member who was either doing the work of
the ill member or caring for that member.

Crop income increases over the 1997 to 2000 period also. The finding that
households suffering fi'om an adult death had a smaller increase in their crop income than
do households not reporting an adult death is consistent with prior expectations.

Finally remittances do fall in line with a priori expectations. The level of remittance
income has declined for all households between 1997 and 2000. But the decline is
especially great for households incurring an adult death between 1997 and 2000
according to table 8. An important question to ask at this point is what role does adult
death play in the decrease in remittances. From table 2 we saw that the deceased
members were not earning a large value of remittance income. Therefore the loss of a
deceased adult’s remittance income is not the reason for the dramatic decrease in
remittances between the two surveys. The loss of remittance income appears to reflect
the reallocation of household labor and the return of household members from off the
farm back onto the farm, as shown in Table 12. One important impression that emerges
from these findings is that a death in the family may result in family members working
off the farm being brought back onto the farm to handle the on-farm labor activities of the
deceased member. This might reflect a priority on food production objectives over cash,
or it might reflect the household’s need for someone to handle other kinds of tasks not

measured in the survey that the deceased member was formerly handling.

25

Table 7. Changes in Farm Characteristics Between 1997 and 2000, stratified by
whether households incurred an adult death in the interim period.

household not incurring a death household incurring an adult

 

 

(1323) death (89)

1997 2000 Mean 1997 2000 Mean w/death-

Mean Mean change Mean Mean change w/out
Variable (2000-1997) (2000-1997) (t-stat)
Area 4.25 5.15 0.90 3.75 5.26 1.51 0.61
Cultivated (0.9822)
(hectares)
Asset 95,618 81,061 -14,555 61,537 52,888 -8649 5609
Value (0.2599)
(Ksh)
Crop 59,023 70,442 1 1,419 48,201 54,244 6043 -5 376
Income (0.4979)
(Ksh)
Crop 16,560 17,433 873 13,660 13,824 164 -709
Income/Ha (0.2328)
(Ksh)
Maize 1,196 1,566 370 787 984 198 -172
Production (0.5 1 86)
(KG)
Livestock 36,701 33,260 -3,441 29,623 25,564 -4060 -619
Income (0.0869)
(Ksh)
Total hh 211,631 190,369 -21,262 205,595 133,642 -71953 -50691
income (1.5281)

 

 

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegemeo Institute, Nairobi, Kenya.

26

Table 8. Changes in Off-farm Activities and Income from 1997 to 2000, stratified by
Whether the Household Incurred an Adult Death

 

 

household not incurring a death household incurring an adult death
(1323) (89)

1997 2000 Mean change 1997 2000 Mean change w/death-

mean mean (2000-1997) mean mean (2000-1997) w/out
Variable (t-stat)
Off-fann 1.21 1.59 0.38 1.18 1.52 0.34 ~0.04
employment (0.285)
(members per
1111)
Remittance 9,875 5,072 -4,803 1 1,772 2,909 -8,863 -4,060
income (Ksh) (-1 ’05)
Off-farm 90,020 63 ,009 -27,01 1 104,795 39,191 -65,604 -38,593
income(Ksh) (1.358)
Farm Wage 2,401 2,026 -375 781 1,509 728 1,103
labor income (0.549)
(Ksh)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegemeo Institute, Nairobi, Kenya.

Finally the next three tables explore the changes in household make-up in more detail.
As has been found in other studies including the Kageara study (World Bank, 1997), the
dependency ratio actually decreases in household experiencing an adult death. In this
study, the dependence ratio does decrease less than the dependency ratio decreases in
households not suffering from the loss of an adult (Table 9). New members were
identified in the 2000 survey. New adult members are a combination of members who
lived on farm less than 5 months in 1997 and then increased their time on farm to more
than 5 months by 2000 and members who are not recorded at all in the 1997 survey but
are included in the 2000 survey. Table 10 shows that most new male members are
children of the head of the household they are entering. However, most new female

members are in the “other” category. This category includes both relatives and non-

27

relatives that are not given a specific category.

Tables 11 and 12 look further at the income characteristics of the new adult members.
Because the category “new people” is made up of those members in 2000 who were not
included in the 1997 survey at all, income measures are not available for them.
Therefore, table 11 gives the characteristics for only the new members who were
enumerated and earned income in the 1997 survey. Theses members are called new
members because they spent less than 5 months on farm in the 1997 survey but more than
5 months on farm in the 2000 survey. The two tables show that new members are
important to the income eamings of their households in 2000. One interesting result of a
close look at the new members is that while none of the 44 new members in the Coast
and Western provinces earned income in 1997, all of them are new members of
households incurring an adult death. This shows conclusively that households incurring
adult death do increase their adult membership as part of the coping process.

Table 9. Dependency Ratios (#Children(<15 years old)/#Adu1ts(>15 years old))

Ratio in 1997 Ratio in 2000 Change
Households w/death 1.27 1.04 -0.23
Households w/out death 1.56 1.05 -0.51

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegemeo Institute, Nairobi, Kenya.

28

Table 10. Characteristics of New Members by Gender 2000

 

 

Total # of Age Years of School Relationship to Household Head (%)
new members (ave) (ave) head spouse parent child other
Male 43 28.56 8.09 6.98 2.33 0 60.47 30.23
Female 36 30.28 7.64 0 22.22 ‘ 0 33.33 44.44
All 79 29.34 7.89 3.79 11.39 0 48.10 36.71

Households not Incurring Adult Death

Male 692 31.07 7.19 22.11 0.87 1.16 47.54 28.32
Female 585 31.90 6.40 2.05 25.81 5.30 42.05 24.79
All 1277 31.45 6.83 12.92 12.29 3.05 45.03 26.70

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegenreo Institute, Nairobi, Kenya.

Table 11. Characteristics of New Members Reporting Non-farm Income in 2000 by
Gender

 

Total # # of new Age Years Off Farm

of new members (ave) of Income in
members deriving School 2000 (Ksh) Relationship to Household Head (%)
income (ave) head spouse parent child other

 

Households Incurring Adult Death (Percent of members reporting zero ofl" farm Income in 2000: Male-96. 65,
Female-90. 82, All-94.58)

Male 43 6 32.83 10.50 50200 33.33 0 0 66.67 0
Female 36 9 32.67 6.78 44078 0 33.33 0 33.33 33.33
All 79 15 32.73 8.27 46527 13.33 20.00 0 46.67 20.00

Households not Incurring Adult Death (Percent of members reporting zero ojf farm Income in 2000: Male-3 .35,
Female-9.18, All-5.42)

Male 692 173 39.57 8.22 58957 49.71 2.31 0.58 39.31 8.09
Female 585 89 36.67 6.78 32432 8.89 35.56 3.33 31.11 21.11
All 1277 262 38.57 7.73 49947 35.74 13.69 1.52 36.50 12.55

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegemeo Institute, Nairobi, Kenya.

29

Table 12. Characteristics of New Members (differentiating between those members
newly entering the household and therefore not included in the 1997 survey, new people,
and those who have increased their time on farm from <5 months in 1997 to >=5 months
in the 2000 survey, extra members)

 

 

 

Number of Mean Years of % of new Informal Farm Remittance Salary
Members Age School members Income Wages and from
in the deriving from from 1997 Pension 1997
category income in 1997 from 1997
1997
New People
Male 559 30.43 6.51 n/a n/a n/a n/a n/a
Female 532 32.56 5.97 n/a n/a n/a n/a n/a
All 1,091 3 1 .47 6.24 n/a n/a n/a n/a n/a

 

Extra Members (Income is reported for only those members reporting that category of income in I 997 *)

 

Male 176 32.49 9.60 43.18 170,896 0 22,470 61,191
Female 89 27.28 9.48 24.72 18,800 0 240,000 56,057
All 265 30.74 9.56 36.98 166,555 0 42,245 59,980

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegerrreo Institute, Nairobi, Kenya.

* Percent of members reporting zero values for income by category: Salary: Male-61.36, Female-76.40,
All-66.42. Informal Income: Male-93.75, Female-98.90, All-95.47. Remittance and Pension: Male-94.32,
Female-98.88, All-95.85.

Farm level effects of disease-related adult death have and will continue to be
important to agricultural research. The possibility that productivity is already at a
weakened state from continued morbidity conditions related to common diseases in SSA
shouldn’t be overlooked in the literature. This is why, while this paper examines the
impacts of disease -related deaths on a series of household and farm characteristics, it also
emphasizes the points in the data which indicate already compromised status, especially

at the farm level.

30

4.0 ANALYTICAL FRAMEWORK

This section outlines the framework for our empirical analysis. The data sets used for
this analysis do not record the specific disease causing an adult death. Rather they record
“disease” as a general cause of death. Therefore, this study analyzes the effects of
disease-related deaths at the household level.

The costs of adult death on households can be both direct and indirect. Direct costs
are those associated with caring for a terminally ill individual, if death was preceded by a
long illness, as well as paying expenses associated with funerals and burial. Indirect costs
are associated with losing the income generated by a productive adult as well as their
embodied human capital and labor potential. This paper will treat disease-related
economic impacts as loss of labor and assets and look for evidence of loss in income and
demographic characteristics.
4.1 Household Model

In this section we address the question: How does adult death affect household
decision making and behavior? First we need to understand the factors which affect the
options available to households. If income generating activities are chosen at the
household level then household decisions will be affected by household resource
constraints. The approach used here is to assume that a household maximizes its utility
function for each given production period. In a unitary model of household decision
making there are several constraints facing utility maximizing behavior. There is an
income constraint, time constraint, as well as a production constraint (Singh, Squire and

Strauss, 1986). These constraints affect productivity and have many implications for this

-31-

study.

Each household will allocate its resources given a set of objectives for leisure, market
goods, non-marketable goods, and future productivity (Singh, Squire and Strauss, 1986
and Haddad et a1, 1997). If the household’s budget constraint is not binding, and where
labor markets function, households will hire labor when they find the marginal value
product of additional labor on their crops exceeding the agricultural wage rate. However,
the death of a productive adult may simultaneously increase the need for hired labor and
reduce the household’s ability to afford it. _

Several types of deaths are typically preceded by a long illness. As with any terminal
illness, economic impacts at the household level will be felt strongly while an individual
is still alive (Beegle, 1997). For those households incurring a death shortly after the 1997
survey, it is quite possible that the impacts will not be measurable between the 1997 and
2000 surveys, because the effects of a terminal illness would have already been measured
in many of the 1997 data. Figure 1 introduces alternate time lines associated with illness
and death over the period of the two surveys used in this paper. The first shows the
possibility that disease onset happened before the 1997 survey was conducted, while on
the second time line disease is shown to have come on after the survey was implemented

in 1997.

-32-

Figure 1. Timing of Illness and Death

 

 

 

 

 

 

Survey I Survey II
Disease Death
Onset
1
Survey I Survey 11
Disease Death
lOnset

 

 

 

 

 

 

Depending on which of the graphs in figure 1 is actually the case, the results of any
analysis done using this data without morbidity information will be different. As there
are no morbidity indicators in the data it may will be difficult to control for some impacts
that will have been felt prior to the death of a household member. The timing of illness
and death in relation to the timing of the 1997 and 2000 surveys is raised to draw the
readers’ attention to the potential limitations of measuring the effects of illness and death.

Table 13 also shows that both scenarios are completely plausible.

-33-

Table 13 Distribution of Deaths by Year of Death

 

 

1997 1998 1999 2000 Total number of
Deaths
Number of 1 1 29 30 19 89
Deaths
Percentage of 12.36 32.58 33.71 21.35 100
total deaths

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and
2000, Tegemeo Institute, Nairobi, Kenya.

4.2 Household Framework
4.2.1 Effects on Household Size

Household demographics are expected to change when adult death occurs. The
specific effects of a terminal illness ending in death and an unexpected death may differ
with respect to many household level outcomes. However, data is not available on the
duration of illness before death, and therefore my analysis focuses on how households
that have incurred an adult death in the family affects the household’s demographic
features, specifically. For household demographics this paper will focus on death’s effect
on the number of children and adults within a household. The change in the number of
adults is measured after controlling for the adult who died. I hypothesize that the number
of children living in the household will be affected by adult death as shown in figure 2.

Figure 2 Number of Children living in a Household

Disease onset Death

 

 

 

 

 

 

-34-

A household experiencing an increased drain on its resources, such as a terminal
illness, may decide to send unproductive members to live elsewhere. As net consumers
of household resources, children may be the first members to be sent to live with other
relatives. Thus I hypothesize that the number of children in a household is expected to
decrease in the event of an adult death.

The additional needs of care giving during illness and other related household tasks
after death may cause the number of adult members to increase due to an adult death (not
counting the adult who died). A household with an ill member rrright find itself
allocating more labor to care-giving than it expected. The need would be increased by the
presence of a terminally ill individual. This would indeed drain the labor resources of the
household. Figure 3 shows that while a household member is ailing the care giving
requirement is expected to be higher than before the member fell ill. I hypothesize that
the number of care givers would then drop after the member has passed away as shown in
figure 3.

Figure 3 Caregiving Needs

Disease Onset Death

 

 

 

 

 

 

-35-

At the same time household use of farrrily and hired labor for farm and other income
generating activities is hypothesized to decline during an illness because of drains on time
having to do with care giving as well as redistributing tasks of members otherwise
engaged in farm or wage labor. After the occurrence of an adult death the availability of
such labor is expected to decline. The changes in patterns of household use of family and
hired labor are dependent on the completeness of labor markets and the budgetary
limitations of a household, as well as the marginal product of farm 1abor(MPLf) and the
market wage(wm). There are three proposed scenarios:

Figure 4 Sum of Hired and Family Labor

AIDS Death

 

 

 

 

 

 

In the first scenario farms are operating under well-functioning labor and credit
markets such that there is no difference in any household’s use of labor due to adult death.
It is only necessary for these households to hire in labor, which is assumed to be available,
to compensate for the labor lost from the death of the adult member. However, if budgets
are constrained, households may find that they cannot pay the actual marginal productivity
for hired laborers, so the second two scenarios may be more plausible. In the second

scenario there are two alternatives, 2a) labor markets are imperfect and additional family

-36-

members are not available to compensate for the loss of a productive member, and 2b)
labor markets are imperfect, additional family members are available, but MPLf < Wm. In
these alternatives the same result is expected,4 a reduction in the number of laborers used
on farm. The difference in number of men in a household is one way to estimate the
effects of adult death on family labor.

The third scenario is such that labor markets are imperfect, additional family members
are available and MPLf > Wm. In this situation these members will work on farm rather
than to engage in off-farm employment. Likewise a decision tree could be constructed to

demonstrate what may happen to family labor specifically in scenarios two and three.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Adult Death Occurs
I F arnily member wm> MPLf F arnily member wm<MPLf
l Member remains off-farm Member called home
No change in # of off-farm Farm activities remain stable,
laborers, stable remittance remittance down, # of off-
income farm workers decreases

 

 

 

 

 

 

 

‘Noted here is the common practice of having household or family members off-farm working and
remitting income back to the farm. Those members remitting income to the farm play an important role in
the event of an adult death. As described above, they may indeed come back to the farm in order to further
help the family to continue current farm practices. They may alternatively stay off-farm to continue
remitting income borne (World Bank, 1999).

-37-

4.2.2 Eflects on Assets

The quantity and value of assets are expected to decline in households incurring a
death in the family. Effects on assets may be difficult to measure, though, for households
in which the death of a farmly member is preceded by a long period of illness. Assets may
be reduced for many reasons relating to death of a family member. Medical expenses and
the loss of any labor income may induce reduction in asset levels. The need for proper
funeral and burial arrangements might firrther reduce household assets. Although, the
year of death of a household member will be very important in quantifying its effects on
household income, there is little way of knowing where the 1997 survey falls in terms of
the time line offered in Figure 5.

Figure 5 Value of Assets

Disease onset Death

 

 

 

 

 

If the enumeration of households in 2000 begins somewhere near the death, it would
be difficult to measure the effect. Ifthe enumeration began in 1997 before the onset of an
AIDS related illness, the results of any subsequent analysis may be far more telling of a
pattern of asset reduction. One result of asset level reduction is the loss of productive

agricultural assets. This could exacerbate any loss of productivity suffered by a farm.

-33-

4.2.3 Eflects on Off—farm Income

Off-farm income can be separated into three distinct portions for this framework:
1) Remittance income of other farmly members not living on the farm, 2) off-farm income
of the deceased, and 3) off-farm income of remaining household members living on the
farm. The distinctions are important. The income of the deceased is expected to be lost to
the household. In order to examine household coping mechanisms I compare off-farm
income of surviving household members. Remittance is treated separately as it is not
expected to move in the same direction as other income off-farm variables. Off-farm
income may be conceptualized this way:

Figure 6 Off-Farm Income

Disease Onset Death

 

Income of deceased Zero

 

Rerrrittance income Remittances possibly

 

 

 

increase
Income of remainin Income of others
family members could remain or decreasie t

 

 

Off-farm income has the potential to progress similarly to labor as the result of an
adult death. Ifthe MPLf > wm then off-farm income may go down as family members are
called back to the farm. The death of an adult may actually provide some opportunities for
other adults in the family to now manage the family farm and therefore migrate back to the

farm to assume new responsibilities. On the other hand, if MPLf < wm , then off-farm

-39-

income may remain unchanged as members continue in their off-farm employment.
Remittances, however, may increase if household members choose to stay away from
home instead of returning home because: when a household is in more need there is the
possibility that those sending income to the household will increase their remittances.
4.2.4 Eflects on Farm-Income

Farm income may be affected in several ways. This analysis considers three variables,
1) area cultivated by a household, 2) crop income per hectare, and 3) differences in yields
of specific crops. The only crop examined individually in this paper is maize5 as it is the
most important and widely cultivated food crop in Kenya, and allows for within-district
comparisons for almost all districts. The effects of adult death on farm income are
demonstrated in Figure 7.

Figure 7 Farm Income

Disease onset Death

 

 

 

 

 

 

Many of the same arguments used in subsection 4.2.2 will apply here. 1) With perfect
labor markets and unconstrained household budgets, it is likely that the area under

cultivation will not change, similarly crop income would not change nor would crop

 

5 A problem of examining the effects of death on maize production is the inability to make gender
differentiations in crop cultivation patterns. The death of a male, for example, may decrease cash crop

production, but increase food crop production

-40-

yields. 2) With a less than optimal amount of labor, it is still possible that area under
cultivation would not change, although yield may decrease with lower labor inputs. 3) It is
also possible that area under cultivation would decline as a result of reduced labor inputs.
In this case, output per hectare may not go down, but overall yield and therefore income
would be reduced. Composition of crops cultivated on farm may also change to reflect
labor saving strategies being implemented. These strategies may further hurt farm income.

Chapter 5 will explore the hypotheses presented conceptually here, empirically.

~41-

5.0 EMPIRICAL SPECIFICATION

This section presents the empirical methodology for analysis. Variables will be
included in this study predominantly as differences between 1997 and 2000 survey values.
Differencing the data is designed to control for time-invariant unobserved variables and
therefore eliminate some of the orrritted variable problems that commonly introduce bias in
cross-sectional analysis.

For estimation purposes the sample has been aggregated from the individual to the
household level, such that all interpretations will be at the household level. It is important
to clarify here that most information was available only at the household level, such that
aggregation was a necessary step. Individual data was available for non-farm income
measures as well as personal characteristics such as education and age. Individual
contributions to farm labor were not available, such that all on-farm variables were taken
from the household level. The individual information that is included in the estimation
process includes the personal characteristics of the deceased household member. The
empirical model can be reduced from the following two equations:

(1) Iii/997.; = “0.1997 + “IX: + “221997 + (13$; + 8i

(2) ”.2000; = 702000 + 1’1Xt + 7222000 + 7321)} + 31

Where i is the household and j is the district, Yi is a vector of endogenous, time-variant,
household-level variables, 0.0 is a constant, X, is a vector of characteristics of the deceased
individual of the household, Z is a vector of time-invariant variables like soil quality, agro-
ecological profile, access to infrastructure, etc.; D is a vector of district dummy variables,

and a is the error term assumed to be normally distributed with zero mean.

.42-

Since this paper is comparing specific households across two time periods, the
dependent variables are differences(2000 survey-1997survey). Therefore the regressions
will be carried out by subtracting equation 1 from equation 2,

(3) (Yuoooj - Yum) = 00,2000 - (10,1997) + (71 - a1)(Xt,2000 - “997) + (72 - a233,)

The time-invariant variables have dropped out of the equation as they are not expected to
vary over a three year period. This will eliminate potential bias that would occur in cross-
sectional data from the omission of relevant time-invariant variables. Equation 3 will
reduce to:

(4) (YT-.2000.) ' Iii/997;) = (751-) + ,B,(X,-)+ flZZDj*1 + 5':

Because the X vector is characteristics of the deceased members, they only have
observations for 1997 and hence this vector of variables enters the model in levels, not
differences, as shown in equation 4.

5.1 Description of Variables

5. 1.1 Independent Variables

The X vector includes the following individual-specific information:

a. A 0/1 dummy variable for household incurring an adult death between 1997 and 2000;
b. A 0/1 dummy variable specifying the gender of the deceased adult member (=1 if
female);

c. A continuous variable specifying the age of the deceased in 1997 (I also test for non-
linear effects of age of the deceased);

d. A continuous variable specifying the years of education of the deceased;

e. A dummy variable specifying the relationship of the deceased to the head of household

-43-

f. A set of 0/1 dummy variables as to whether the household incurred a death in 1997,
1998, 1999, or 2000.
I enter these variables into the model in the following sequence, and report the results
for each model in appendix 1:
(5) same as equation (4) with the X vector including only the adult death dummy variable;
(6) same as (4) but with the X vector including the adult death dummy and gender of
deceased member dummy;
(7) same as (4) but with the X vector including the adult death dummy, gender of the
deceased, age of the deceased, years of education of the deceased;
(8) same as (4) but with the X vector including the adult death dummy, gender of
deceased, age of deceased member, years of education of deceased member, relationship to
household head dummy, and year of death dummy.
Regressions will also be re-run using the quadratic form of age, with F -tests used to
determine its applicability. The quadratic form of age will only be reported if relevant.
The characteristics of deceased household members are hypothesized to be important
to explaining the possible impacts of an adult death on the household unit. Age and
education level are used here as proxies for human capital. A household would be
expected to suffer more severely from a death if the deceased member was an important
store of human capital. Gender of the deceased member may also elicit different
household responses in bringing in new members or sending others away. The deceased
member’s relationship to the household head is also expected to play a role in the coping

strategies employed by the household, particularly if the member was indeed the

-44-

household head. This variable also plays a role in the human capital assumption. The
household head is expected to make important decisions for the household. The loss of
this person may be particularly difficult to cope with. To ameliorate the problems of
lacking morbidity data, the year of death is included as a dummy variable. This will help
to control for timing issues in the dependent variables.
5.1.2 Dependent Variables
Household composition

There are three household composition variables used in this analysis. They are
differences in number of women, men and children in a household between the 1997 and
2000 surveys. The differences exclude the effect of the loss of a member due to death,
such that the estimation looks at the size of a household excluding the deceased member.
A priori expectations of these variables were discussed in Chapter 3, however, it is
reinforced here as depending on the timing of the 1997 survey between onset of illness and
actual death of a family member. The number of children is expected to decrease in order
to reduce the number of net consumers in the household; however, this hypothesis may be
reversed if the number of adult household members increased sufficiently to accomodate
necessary household firnctioning. The number of men and women are expected to
increase if the household is not able to hire labor to compensate for the loss of an adult and
to compensate for the extra labor needed care for an ill adult.
On-farm efl'ects

The variables for farm level effects are differences in value of assets, total crop income

and crop income per hectare (the calculation of on-farm income was discussed in chapter

-45-

3), area under cultivation, and maize harvests in 90 kg bags. Depending on the location of
the survey on the time line fiom figure 1, the effects of an AIDS related adult death may
vary. Ifthere is compensation within the household as per labor allocation, there may be
negligible effects of adult death on income, cultivated area, and harvests. Assets are
expected to be liquidated during an illness such that the placement of the 1997 survey may
fail to catch a large portion of the actual effect if the onset of illness preceded the 1997
survey. Again, lack of data on morbidity in the data set inhibits effective measurement of
the household impact of an AIDS-related death, reflecting the more general effect of an
adult death.
Off-farm eflects

The change in both off-farm income and number of family members living on farm
and engaging in off-farm employment will be analyzed. Off-farm income will be
separated into three different variables, 1) difference in off-farm income, this will include
labor income and small-medium scale enterprises, 2) difference in farm labor income only,
and 3) differences in the value of remittances. The calculation of these income variables
are further discussed in chapter 3.

Previous literature finds that remittances will be decreased in the event of an adult
death, all other effects held constant (Ainsworth, 1998). This study hypothesizes that
while income is expected to go down as a result of adult death, remittance could either
stay the same or increase, depending on the marginal productivity of household members
in off-farm employment relative to expected returns from reallocating that person back to

the farm (which may also involve a combination of farm and off-farm employment, but

-46-

based from the rural domicile rather than from different employment opportunities
elsewhere). With greater need on farm, members who live away may send more money
home than usual or at least not decrease their contribution.

The last off-farm dependent variable is number of household members engaged in
off-farm employment. This variable, like the demographic variables, omits the effect of
the deceased member. Therefore, the number of household members engaging in off-farm
employment in 1997 was calculated to exclude the deceased member, and then subtracted
from the 2000 figure.

District level dummy variables

To control for effects on the dependent variables district level dummy variables were
included. The district experiencing neutral effects of the dependent variables were
dropped from the analysis. When analysis is presented on the district variables, the

comparison to an average district will make the analysis consistent.

-47-

Table 14. Definition of Variables and Descriptive Statistics

 

 

 

 

Variable Definition Mean of Std. Dev.
differenced
variable
Dependent Variables (Y i)
off-farm Change in all labor income, formal and informal, -29,334 253,976
income including small-meditun scale enterprises
(these numbers are gross and overestimated)
farm wage Change in labor income, from farm related work -3 09 18,001
income
off-farm Change in number of household members engaged in 0.3746 1.2517
employment off-farm employment
remittance Change in value of remittances -5,047 34,178
farmed land Change in area(ha) under cultivation 0.9377 5.5480
assets Change in the value of a household’s assets -14,201 203,080
maize harvest Change in maize harvests in kgs 360 2,981
women Change in the number of women in a household, 0.2238 1.2063
excluding the effect of the loss of a female member to
death
children Change in the number of children in a household -.2592 2.1964
men Change in the number of men in a household, .1941 1.3424
excluding the effect of losing a male member to death
livestock Change in livestock income, including sale of livestock -3,478 63,646
income and livestock prices
crop income Changes in both crop income and crop income per 11,095 96,475
hectare will be used
crop income This will not be done by crop, as area planted to a 831 27,210
per hectare specific crop is not included in the data
household change in total household income, includes crop -24,313 296,637
income income, livestock income, remittances and pension,
non-farm income, including farm wage income.
Categorical Explanatory Variables Frequency % households
(number of incuning a death
households reporting a 1
reporting a 1)
adult deaths Equals 1 if a household has reported any deaths on the 85 100
2000 survey, 0 otherwise
head deceased Equals 1 if the household head is dead, 0 otherwise 24 28.23
spouse Equals 1 if the household head’s spouse is dead, 0 13 15.30
deceased otherwise

 

-43-

other relative

Equals 1 if the relationship of the deceased member to

12 14.12

 

 

 

deceased the household head is other than spouse or actual head,
0 other wise
sex of This is a dummy variable taking a value of 1 if the 39 45.90
deceased deceased member is female,
0 if the deceased member is male 46 54.12
year of death This is a dummy variable taking a value of 1 if the 1 1 12.94
1997 death occurred in 1997, 0 otherwise
1998 takes a value of 1 if the death occurred in 1998, 0 26 30.60
otherwise
1999 takes a value of 1 if the death occurred in 1999, 0 29 34.12
otherwise
2000 takes a value of 1 if the death occurred in 2000, 0 19 22.35
otherwise
Variables in Levels: Mean Standard Deviation
education Indicates the level of education attained by the 6.60 4.09
level deceased, if the household has reported more than one
death it equals the highest level of education among the
deceased
deceased age This variable indicates the age of the deceased member 37.53 12.51
deceased age The age of the deceased member squared, accounts for 1,563 943
squared non-linearity in the age term
district Dummy variables for province and district of residence included in all regressions
durmnies

 

-49-

6.0 RESULTS

This section presents the results of the econometric estimation discussed in section 5.
There are 4 separate estimations for each dependent variable. They all include district
durruny variables as well as the independent variables listed in section five under points 5-
8. The contents of the alternative X vector for each estimation are:
(1) X1 vector including only the adult death dummy variable;
(2) X2 vector including the adult death dummy and gender of deceased member dummy;
(3) X3 vector including the adult death dummy, gender of the deceased, age of the
deceased, years of education of the deceased;
(4) X4 vector including the adult death dummy, gender of deceased, age of deceased
member, years of education of deceased member, relationship to household head dummy,
and year of death dummy.
Discussion will focus on the results significant at least at the 10% level. This section will
be organized according to the following categories of the dependent variable: household
demographics, off farm effects and on-farm effects. The first part of each subsection will
begin by presenting some of the general predictions. Further sub-sections will discuss
significant results in more detail. Results in tables 15-17 highlight some significant results
from the regressions. A comprehensive set of regression results is available for all

regressions in Appendix 1.

-50-

6.1 Household Demographics

The effects of an adult death on household composition are the most interesting of this
analysis.6 The results are consistent with a priori expectations as well as some of the
findings of earlier research presented in the literature review. One key hypothesis was that
an adult death in the family would induce other non-resident adults in the family to return
to the farm because of increased demands for labor within a household. Number of men
and women were estimated separately because the potential effects of an adult death,
depending on the individual characteristics of the deceased household member, may be
different for men and women.

Table 15 presents the predicted effects of on the change in the number of household
members resulting from an adult death. The predicted response is sensitive to the
deceased’s gender, age, and level of education. For example, the predicted effect of the
death of a 35 year old adult male with 4 years of education on the change in number men
in a household is 0.3 8, all else held constant. This prediction is statistically significant at
the 5% level as are the predictions shown in table 15 for effects of death on change in
number of children. The predictions listed for change in number of women are not
significant and are therefore not listed in table 15, however, the analysis will refer to the
regression results from Appendix 1: table 1 to discuss significance in the regressions using

change in number of women as the dependent variable.

 

“Please note that the dependent variables were constructed to orrrit the deceased individual in the
1997 totals for both number of men and women.

-51-

Table 15. Predicted Effects of Adult Death on Household Composition

 

Predicted Household Behavioral Rgsmnse

 

Household Member who A # of Women A # of Men A # of Children
died from disease

Male aged 35, edu: 4yrs -0.2236 0.38047 0.922

Male aged 35, edu: 9yrs -0.1 166 0.5579 0.8195

Male aged 5 5, edu: 4yrs -0.226 -0.3656 41.468

Male aged 55, edu: 9yrs -0.119 -0.1881 41.5705
Female aged 35, edu: 4yrs 0.3495 -0.6275 0.2965

Female aged 35, edu: 9yrs 0.4565 -0.45 0.194

Female aged 55, edu: 4yrs 0.3471 4.3735 -1.0935
Female aged 55, edu: 9yrs 0.4541 -l.l96 4.196

 

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

6.1.1 Men

Results reported in Table 15 are predictions based on the combination of age,

gender and education of the deceased household member. In the case of change in number

of men, these variables are jointly significant at the 5% level.

Education

The general results are interesting across all characteristics of deceased members.

As can be seen from table 15, when moving from a deceased member of education 4 years

to education 9 years, the effect on the number of men in a household becomes either a

larger negative or a smaller positive. This supports the possibility that a member with

more education may also spend less time on farm, therefore not requiring replacement

from the production side.

 

7Results significant at least at the 10% level of significance are bold in tables 15-17.

Age

With the death of an older woman a household is less likely to increase its number
of men than with the death of younger women or men. This may show that the death of a
younger person, especially a younger man induces a household to replace adults with men
in greater proportion because of family relationships. Tasks carried out by elderly
members would be handed over to younger family members, so there is no major labor
shock when an elderly person dies. Replacing young adults who have died may manage to
preserve the social and familial integrity of the household.

Gender

The death of a fifty-five year old male member has a negative impact on the
number of male household members (-0.51). The reasons for these differences may indeed
be because of age. The older member may carry a lesser work load such that the death of
an older member may not negatively effect the labor supply within the household. If the
older member was ill for some time leading up to his death, he may have required more
care than any younger members, requiring the extra attention of female members. Upon
his death, the members caring for him can take up more farm activities, freeing men to
pursue off-farm activities.

The effects of the death of a woman on the number of male household members
appears to be negative at all three test ages. The death of a thirty-five year old woman
(-0.19), and a fifty-five year old woman (-1.37), both of education 4 years, show that the
death of a woman does not increase the demand for male labor within the household. The

death of the woman may also lead to husbands or sons losing rights to land in matrilineal

-53-

societies, and so going elsewhere. In patrilineal areas, the death of the woman may induce
another caregiver (sister or cousin of the deceased wife) to come into the family while the
man goes to work in the town for awhile. Rather the death of these women at different
levels and in keeping with their age may indicate the need for fundamental changes in the
family structure, including the possibility of men then to leave the household in search of
other occupation following the death of a woman. The death of an older woman may
induce some male members to leave, especially older male members. They may leave to
join households of older children. The exit of older adult members does not indicate the
dissolution of the household, as the older members may not make up the core of the
household.
6.1.2 Women

The general results relating to the change in number of female adult household
members are similar to those reported in 6.1.1 in that the more education the deceased
individual has the smaller the negative impact on the number of adult women in the
household. It seems very possible that in some cases, it is in the best interest of the
household both for productive concerns, food security and even perhaps community
relations to send female members away following the death of a member because of their
potential worth to the household in bride prices or for any earnings they may get as house-
help or other such employment.
Gender and relationship to household head

In the fourth specification of the regression both gender of the deceased member

and the spouse of the household head dummy variable were significant (Appendix 1: table

-54-

1). The regression results show that, all else remaining constant, if the deceased member
is a woman and the spouse of the household head then the effect of her death on the
change of female household members is (-O.56). For a woman who is not the head’s
spouse the effect is (0.57); for a man who is the spouse of the household head the effect is
(-1.13). These results may not be all that surprising as the death of a male spouse may
induce female adult members to move to other households. Again this does not
necessarily mean the household will dissolve; there are typically more than one male and
female adults residing in the household, and these other members may continue to hold
the household together. The death of a female spouse may also induce other female
members to seek other opportunities, or flee in areas where staying would imply becoming
the man’s wife.
Age

The third regression using this dependent variable found that the age of the
deceased member was significant (Appendix 1: table 1). The death of sixteen year old has
a smaller impact (-0.31) than do the deaths of older adults and similarly a thirty-five year
old has a smaller impact (-0.68) on the number of women in a household than does the
death of a fifty-five year old member (-1 .07). These results are somewhat similar to those
from the regression using change in men as the dependent variable. A household unit may
be more likely to replace younger members than older members because of their potential
productivity over time. It may also be easier to replace younger people than older people
because of the roles they play in the household and the ease with which a younger person

may be called home by older people.

-55-

6.1.3 Children

Change in the number of children resulting from an adult death did not coincide
with a priori expectations of a decrease in the number of children. Results presented in
Table 15 for change in number of children are jointly significant for the characteristics of
the deceased adult presented at the 5% level. These results were contrary to the changes in
other household composition shown in table 5. Where number of men and women
decreased slightly in response to an adult death, the number of children increased. The
results of the regressions mirror this difference.
Age

As Table 15 shows, the younger the deceased adult the more likely the household
is to increase its number of children. As of yet the emotional ties to the deceased members
have not been discussed. In this case, especially when considering the death of a younger
member, additional children in the household following a death may have something to do
with pain involved with the loss of a child, albeit an adult-child. Continuing a similar
family balance of young and old may also be a priority to families. Furthermore, as the
stronger positive effect of the death of a young male as opposed to that of a young female
may indicate a balance of labor between young men and women.
Education

Education also seems to have interesting effects on the number of children in a
household following an adult death. From Table 15 it can be seen that the higher the
education of the individual, regardless of other characteristics, lower the increase in

number of children. For younger men and women this effect manifests itself as a smaller

-56-

positive change, while for the death of an older person it shows a larger negative change.
This result may show that households with individuals at a higher level of education may
see the importance of either family planning or relocating children in a crisis situation.
6.2 Off Farm Effects

As is shown in table 16 the predicted values of specified characteristics of
deceased members are not as consistent with a priori expectations as they were in the
household composition simulations.

Table 16. Predicted Effects of Adult Death on Off-farm Household Indicators

 

Predicted Household Behavioral Resmnse

 

Household Member who A # of hh members A in Value of A in Value of Off- A in Value of Farm
died from disease engaged in off-farm Remittances Farm Income Wage Income
employment (Ksh) (Ksh) (Ksh)

Male aged 35, edu: 4yrs -0.2571 -372 -19939 -225.656
Male aged 35, edu: 9yrs 0.0369 -822 ~349l4 -248.976
Male aged 55, edu: 4yrs -0.8151 -6472 -81599 1534.344
Male aged 55, edu: 9yrs 41.5211 -6922 -96574 1511.024
Female aged 35, edu: 4yrs -0.6357 3370 -92252 2484.344
Female aged 35, edu: 9yrs -0.3417 2920 -107227 2461.024
Female aged 55, edu: 4yrs -1.1937 -2730 -153912 4244.344
Female aged 55, edu: 9yrs -0.8997 -3180 -168887 4221.024

 

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

6.2.1 Ojf farm labor

The predictions reported for change in number of family members engaged in off-
farrn labor are statistically significant at the 10% level. Regression results do not include
the off-farm labor hours of the deceased individual.
Age

-57-

Generally the more negative impacts on change in members engaged in off-farm
employment come when the deceased members are older members. This could be due to
the important role that older members play in the household’s everyday activities as well
as the household’s social fabric. The loss of older members may signal the need for others
to take over the care of young children. Since the regressions done on household
composition indicate that the death of an older member does not positively effect the
number of adults, then it stands to reason that instead of increasing the number of adults
on farm, a household may just require the remaining adults to spend more of their time on
farm.

Education

As table 16 shows, the more education the deceased member has the smaller the
negative impact or larger the positive impact on the members engaged in off-farm
employment. These results may say something about the education of the other household
members or it may indicate the perceived value of household members’ time spent on-
farm. It is possible that the deceased member with more education spent less time on
farm, such that replacing their income is better done off-farm. This may be especially true
in the case of the young member with a large number of years of education. This
household member is likely to have spent a good deal of his or her time off farm anyway,
such that sending another member off-farm for employment in their absence doesn’t

necessarily take away from existing family labor availability.

-53-

6.2.2 Remittances
Gender

Table 16 predicts that the impact of various sets of characteristics of a male death
in the household will increase the value of remittances sent to the household. However, if
the deceased member is a woman, her death will decrease the value of remittances
received by the household. An observation of Kenyan culture is that often women are sent
to other areas as house-help with the agreement that they will send their earnings home.
The death of a female member may require the return to the household of those women
off-farm and remitting income. Therefore, this prediction seems to keep with a priori
expectations as well as observed family structure. The opposite effect of the death of
males on remittances may indicate that with the loss of a male, and a strong tendency for
males to join the household, women are indeed seeking other opportunities for outside
employment.
Adult death

The effect of adult death on remittance income was hypothesized in section 4 to
cause no change or perhaps an increase in the level of remittances to the suffering
household. However, the only regression using change in value of remittances received by
a household that had significant results was the first regression (Appendix 1: table 8)
which uses only the adult death dummy and district dummies as explanatory variables.
This regression shows a negative impact of adult death on remittances (-6441). This may
be in part due to the loss of any remittance income from the deceased member or the return

of other household members who had been earning remittance income to the household.

-59-

The results of changes in household composition are consistent with this finding. As
members previously working off the farm are called back to the farm in response to
another adult member’s death, the income that they might have been remitting dries up
when they migrate back.
6.3 On Farm Effects

An interesting outcome in Table 17 is that for every prediction- except that
associated with crop income per hectare, the value of the prediction is negative. The
special role of older women in a household has already been touched upon in earlier
discussion. It is highlighted again here as many of the predicted effects of the death of an
older woman for on-farm indicators are negative. Women are understood to do a large
portion of the farm work such that the loss of a woman may be more devastating to crop
income and maize harvests. The cause of negative impacts on total household income
may stem from the effects of the death of an older woman on household composition and
off-farm income indicators. If a younger woman comes home to replace an older women,
and off-farm income decreases, the loss of the older woman and her farm experience
causes crop income to decrease, then total household income may also decrease.
6.3.1 Crop Income and Crop Income per Hectare
Education

The loss of a well educated man may reduce the farm management capacity, or the
farming expertise of a household. Therefore the loss of such human capital could
justifiably reduce the productivity per hectare of land. The same result does not hold for a

well educated woman, although her loss does increase crop income per hectare less than

-60-

the loss of a less educated woman.
Gender

While table 6 shows that crop income per hectare increases between 1997 and
2000, it is not what was hypothesized in section 4. The second regression using crop
income per hectare as the dependent variable (Appendix 2: table 3) shows that if the
deceased member is a woman, then crop income increases ksh 9772 all else remaining
constant. This result makes less sense when considering the regression on household
composition. The death of a woman in those regressions does not seem to increase the
number of men or the number of women in the household. Perhaps the positive effect of
an adult death on the change in the number of a household’s children has something to do
with this, if the children are working. Also any new members with special skills or trading

businesses may have something to do with this result.

-61-

Table 17. Predicted Effects of Adult Death on Farm Household Level Indicators

Household
Member
who died
from
disease

A # of
hectares
under crops

A value of
Crop
Income

(Ksh)

Predicted Household Behavioral Remnse

A value of
Crop
Income
per
Hectare

(Ksh)

A value of
Livestock
Income

(Ksh)

A in
quantity
Maize
Harvest

A in total
household
asset value
(Ksh)

A in total
household
income

(Ksh)

 

Male aged
35, edu:
4yrs

Male aged
35, edu:
9yrs

Male aged
55, edu:
4yrs

Male aged
55, edu:

9yrs

Female
aged 35,
edu: 4yrs

Female
aged 35,
edu: 9yrs

Female
aged 55,
edu: 4yrs

Female
aged 55,
edu: 9yrs

 

-0.9416

-0.6421

-l.7036

-l.404l

-1.5967

-1 .2972

-2.3587

-2.0592

-11,320

-15,055

-31,960

~35,695

13,505

9,770

-7,l35

-10,870

1,897

-2,353

437

-3,813

12,210

7,960

10,750

6,500

5,728

1,538

2,108

-2,082

-2,06 1

-6,251

-5,681

-9,871

-239

-139

-699

-5 99

74

-486

-3 86

-4,217

-20,637

41,357

-57,777

-32,453

-48,873

-69,593

-86,013

-27,268

-50,713

-l25,688

-l49,l33

-76,815

-100,260

-175,235

-198,680

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

-62-

7.0 CONCLUSIONS

The above analysis shows the households included in the data that are recovering
from an adult death have compensated in large part through changes in household
composition, which have led to decreased remittances and fewer members engaged in off-
farm work. These strategies are completely understandable as households attempt to
continue functioning as close to "normal" as they can. The importance of education in
reducing a household’s need to replace adult members demonstrates what education can
do to reduce dependence on one activity. In Kenya, many rural households have
diversified into areas outside of farming because the education of young members
presented opportunities not available before.

However, with the loss of the intermediate generation, traditional knowledge can
be lost. The ability of households to cope with the loss of external income, and the loss of
expertise in farming or animal husbandry passed from parent to child, will leave them
without formal and informal education. As dependency ratios are increasing in Kenya,
the number of children in the care of the elderly becomes higher and higher. The
environment for education and knowledge transference is strained.

Adult death is a very difficult subject to tackle. A deepening crisis of adult death,
especially one related to such a controversial disease as HIV/AIDS, can make quantitative
work even more difficult. The differing epidemiological patterns of HIV/AIDS compared
to other diseases such as Malaria show a need to look at these diseases separately.
However, finding specific data on HIV/AIDS is very difficult where families do not

necessarily volunteer such information.

-63-

Farm households in particular are of great concern in this field as agriculture is so
crucial to the development food security and stable econorrries in Kenya and in other
developing countries. While an urban bias in many societies can sideline the importance
of investing in the health and well being of the rural sector, more research such as this
shows that a crisis in the rural areas could easily transmit to a crisis in urban areas. Labor
availability as well as food availability are important considerations.

With the understanding that adult death is not decreasing in SSA, relief,
development, and technology development and transfer organizations, and governments
need to ensure that the activities they pursue are mindful of possible labor shortages, and
increased morbidity and mortality. New technologies cannot be labor intensive. Rather,
labor saving technologies will do the most to help households cope with less labor, and to
hopefully continue off-farm enterprise and education.

Continuing the research on off-farm effects such as remittances in households who
have experience an adult death can it shed light on the deepening crisis of adult-death that
is expected in SSA, and may help to explain urban effects as well. If the food security
crisis is observed to be worse for farm families, implications for city families could be
observed concurrently. The possibility that an urban to rural population drain may occur is
another issue, which may be captured in looking at effects on remittances as well as off-
farrn income between households with and without adult death.

NGO’s and government agencies might be able to target their services dependent
on whether the most vulnerable households are those with diverse income bases or mainly

farming households. Implementing services for households consisting primarily of elderly

-64-

and very young members would be a major priority as these are marginal households with
little ability to cope with any further crisis.

To understand more fully the coping mechanisms employed in across small farm
households, tracking families who have experienced adult illness and death over many
years is important. It is also important to track their members and where they go if indeed
they leave the farm for other opportunities. Decisions taken by households thought to be
dissolved may offer lessons on how to preserve household integrity or on what other
options exist for struggling households. Members of dissolved households may give
insights on what is to come for other households as adult death becomes more common.

Concerted efforts in recording morbidity information for rural households in future
research will allow researchers to more accurately characterize the process of coping fi'om
disease onset to death. Loss of assets and drain on labor resources are likely to be felt very
intensely during a terminal illness. Results showing this information will be equally
interesting to researchers and policy makers as the actually results relating to adult death.

Social services and infiastructure to help households as they cope with these
difficulties may be better targeted with research such as what is presented as well as what
is suggested. With a knowledge of areas where the crisis is most developed governments
and other organizations can gauge where information dissemination regarding measures to
be taken for prevention and treatment of HIV/AIDS should be focused. Facilities for care
of orphans, or patients may be needed more acutely in certain areas, especially in Western
Kenya for instance. Communication between various stakeholders in the country should

be facilitated such that efforts towards prevention and education would not be redundant.

-65-

Rather governments, the private sector, NGO’s, and churches, among others, should have
a consistent message and methods.

The above analysis leaves many questions concerning the ability of households to
effectively cope with adversity with traditional mechanism in this special situation.
Ultimately this research is about people who face an already difficult set of factors
regarding choices dictating their health and well being. Trade-offs between healthy and
unhealthy options can be unclear. Diverting the attention of policy makers, donors,
NGO’s or relief organizations from long term sustainable development operations to
HIV/AIDS prevention or treatment would have detrimental effects on other areas of
concern within the developing world. It is, however, crucial to remember that this crisis
factors into people’s lives more and more. Ensuring an HIV/AIDS component in poverty
reduction strategies will allow development to hopefully progress while also trying to slow

the advancement of adult death.

-66-

APPENDICES

-67-

Appendix 1. Regression results

This appendix presents the regression results for all dependent variables except

crop income, crop income per hectare and total household income. The latter three will be

presented in appendix 2. There are four separate versions of the regression reported for

each dependent variable. Please note the following:

1. All regressions include district dummy variables,

2. Coefficients are reported with the absolute value of the corresponding t-statistics

reported in parentheses directly below,

3. Significance is indicated by an asterisk (*=10%, **=5%, ***=l%).

Table 1. Change in Number of Women

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death 0.1426 -0. 1237 0.5553 -0.305
(1.036) (0.671) (0.949) (0.432)
Sex of Deceased 0.5724“I 0.4564 0.5731“
(2.171) (1.629) (1.898)
Age of Deceased -0.0194* -0.0001
(1.739) (-0.008)
Education of 0.0159 0.0234
Deceased (0.477) (0.639)
Deceased -0.2924
Member: (0.706)
Household Head
Spouse of Head -1.1345***
(2.649)
Death Occurred 0.2498
in: 1998 (0.564)
1999 0.1416
(0.320)
2000 0.7958
(1.713)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

-68-

Table 2. Change in Number of Men

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death 0.2937“ 0.6924*** 1.9676*** 1.5439"
(1.917) (3.382) (3.036) (1.967)
Sex of Deceased -0.8568*** -1.075*** -1.008***
(2.925) (3.461) (3.002)
Age of Deceased -0.037*** -0.0372**
(2.995) (2.249)
Education of 0.0334 0.0355
Deceased (0.901) (0.955)
Deceased 0.2095
Member: (0.455)
Household Head
Spouse of Head 0.1229;
(0.258)
Death Occurred 0.0681
in: 1998 (0.138)
1999 0.4244
(0.864)
2000 0.6332
(1.226)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

-69-

Table 3. Changr in Number of Children

 

Regression 1 Regression 2 Regression 3 Regression 4
Adult Death 0.1342 0.4533 2.8965*** 3.4365***
(0.531) (1.34) (2.70) (2.650)
Sex of Deceased -O.6857 -1.1043** -O.6254
(1.416) (2.149) (1.127)
Age of Deceased -0.0551*** -0.0695**
(2.687) (2.541)
Education of -0.0279 -0.0206
Deceased (0.456) (0.334)
Deceased 0.9685
Member: (1 .272)
Household Head
Spouse of Head -0.3768
(0.479)
Death Occurred -0.2776
in: 1998 (0.341)
1999 -1.2267
(1.51 1)
2000 0.0906
(0.106)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,

Tegemeo Institute, Nairobi, Kenya.

-70-

Table 4. Change in Number of Household Members Engaged in Off-farm
Employment

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death -0.0368 0.0613 0.6998 0.4842
(0.260) (0.323) (1.162) (0.664)
Sex of Deceased -O.21 l —0.3 196 -0.3786
(0.776) (1.108) (1.214)
Age of Deceased -0.0261** -0.0279*
(2.273) (1.815)
Education of 0.0603* 0.0588"
Deceased (1.752) (1.702)
Deceased 0.0158
Member: (0.03 7)
Household Head
Spouse of Head 0.4856
(1.098)
Death Occurred 0.0559
in: 1998 (0.122)
1999 0.41 15
(0.902)
2000 0.3707
(0.773)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

-71-

Table 5. Charge in Number of Hectares Cultivated

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death 0.6958 1.2962 0.7281 0.1523
(1.101) (1.531) (0.270) (0.047)
Sex of Deceased -1.2906 -1.l929 -0.6551
(1.065) (0.925) (0.472)
Age of Deceased 0.0079 -0.0381
(0.154) (0.556)
Education of 0.0347 0.0599
Deceased (0.226) (0.390)
Deceased 3.3 143*
Member: (1.793)
Household Head
Spouse of Head -0.0726
(0.037)
Death Occurred -0.5 599
in: 1998 (0.275)
1999 3.549
( 1 .746)
2000 -0.3783
(0.177)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,

Tegemeo Institute, Nairobi, Kenya.

-72-

Table 6. Change in Value of Assets

 

Regression 1 Regression 2 Regression 3 Regression 4
Adult Death 1788 10138 80261 73914
(0.077) (0.327) (0.814) (0.619)
Sex of Deceased -l7950 -29995 -28236
(0.404) (0.635) (0.552)
Age of Deceased -1160 -1857
(0.616) (0.736)
Education of -321 8 -3284
Deceased (0.571) (0.580)
Deceased 24267
Member: (0.246)
Household Head
Spouse of Head 9604
(0.133)
Death Occurred 51492
in: 1998 (0.686)
1999 10580
(0.141)
2000 20048
(0.255)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,

Tegemeo Institute, Nairobi, Kenya.

-73-

Table 7. Change in Kgs of Maize Harvested

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death -73.57 -198 506 486
(0.221) (0.445) (0.357) (0.283)
Sex of Deceased 268 148 213
(0.421) (0.218) (0.290)
Age of Deceased -20.30 -22.88
(0.750) (0.631)
Education of 17.39 19.55
Deceased (0.2 1 5) (0.240)
Deceased 23 1
Member: (0.229)
Household Head
Spouse of Head -321
(0.308)
Death Occurred 194
in: 1998 (0.180)
1999 55.23
(0.051)
2000 -89.35
(0.079)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,

Tegemeo Institute, Nairobi, Kenya.

-74-

Table 8. Change_in Value of Remittances

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death -6441* -6571 -2226 10663
(1.650) (1.256) (0.134) (0.530)
Sex of Deceased 279 -467 3742
(0.037) (0.059) (0.435)
Age of Deceased -72.22 -305
(0.227) (0.719)
Education of -197 -89.53
Deceased (0.208) (0.094)
Deceased 13626
Member: (1.154)
Household Head
Spouse of Head -11436
(0.937)
Death Occurred -4854
in: 1998 (0.384)
1999 -9254
(0.734)
2000 -18606
(1.405)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,

Tegemeo Institute, Nairobi, Kenya.

-75-

Table 9. Change in Value of Off-farm Income

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death -3 222 1 -3 293 1048 15 99946
(1 . 127) (0.086) (0.861) (0.678)
Sex of Deceased -62184 -80734 -72313
(1.134) (1.384) (1.147)
Age of Deceased -2078 -3083
(0.893) (0.991)
Education of -3280 -2995
Deceased (0.474) (0.429)
Deceased 59761
Member: (0.691)
Household Head
Spouse of Head 56835
(0.636)
Death Occurred 43335
in: 1998 (0.468)
1999 61736
(0.669)
2000 15450
(0.159)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,

Tegemeo Institute, Nairobi, Kenya.

-76-

Table 10. Change in Value of Farm Wage Income

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death 1365 1031 -3243 -3287
(0.660) (0.372) (0.368) (0.308)
Sex of Deceased 718 1449 2710
(0.181) (0.344) (0.594)
Age of Deceased 111 88.36
(0.659) (0.392)
Education of -34.22 -4.66
Deceased (0.068) (0.009)
Deceased 2766
Member: (0.442)
Household Head
Spouse of Head -1269
(0.196)
Death Occurred -l409
in: 1998 (0.210)
1999 -854
(0.128)
2000 1 107
(0.158)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

-77-

Table 11. Change in Value of Livestock Income

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death -2113 2207 23022 15415
(0.289) (0.225) (0.73 8) (0.408)
Sex of Deceased -9287 -12863 -7789
(0.662) (0.861) (0.482)
Age of Deceased -344 -181
(0.578) (0.227)
Education of -958 -838
Deceased (0.537) (0.468)
Deceased 2426
Member: (0.109)
Household Head
Spouse of Head -15928
(0.695)
Death Occurred -142
in: 1998 (0.006)
1999 -8055
(0.341)
2000 12957
(0.521)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,

Tegemeo Institute, Nairobi, Kenya.

-78-

Appendix 2. Regression Results

The results presented here are mostly for comparison. Remember that the analysis
uses the regression results associated with constant 1997 prices; note again that all prices
have been inflated to 2000 terms. The comparison is between regression run using the
constant price approach and regressions run using 1997 prices in 1997 and 2000 prices in
2000, again all prices inflated to 2000. Please note:
1. All regressions include district dummy variables,
2. Coefficients are reported with the absolute value of the corresponding t-statistics
reported in parentheses directly below,

3. Significance is indicated by an asterisk (*=10%, **=5%, **"‘==1%).

-79-

Table 1. Chanflin Crop Income (Constant 1997 prices)

 

Regression 1 Regression 2 Regression 3 Regression 4
Adult Death -3322 -1 1623 1 l 142 27787
(0.3 14) (0.820) (0.247) (0.509)
Sex of Deceased 17843 13934 24825
(0.880) (0.646) (1.063)
Age of Deceased -397 -1032
(0.461) (0.896)
Education of -928 -747
Deceased (0.360) (0.289)
Deceased 33869
Member: (1.058)
Household Head
Spouse of Head -53 84
(0.163)
Death Occurred 727
in: 1998 (0.021)
1999 -18739
(0.548)
2000 -6848
(0.191)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

-80-

Table 2. Change in Value Crop Income

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death 609 -6097 15717 30006
(0.058) (0.434) (0.352) (0.555)
Sex of Deceased 14417 10675 21589
(0.718) (0.499) (0.934)
Age of Deceased -427 -1240
(0.501) (1.087)
Education of -620 -463
Deceased (0.243) (0.181)
Deceased 39880
Member: (1.258)
Household Head
Spouse of Head 4089
(0.125)
Death Occurred 8036
in: 1998 (0.237)
1999 -12540
(0.371)
2000 665
(0.019)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,

Tegemeo Institute, Nairobi, Kenya.

-31-

Table 3. Change in Crop Income per Hectare (Constant 1997 prices)

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death -243 -4789 1726 7852
(0.079) (1.164) (0.132) (0.496)
Sex of Deceased 9772* 8646 10313
(1.660) (1.381) (1.523)
Age of Deceased -10.84 -73.36
(0.043) (0.220)
Education of -857 -850
Deceased (1.147) (1.132)
Deceased 2097
Member: (0.226)
Household Head
Spouse of Head -708
(0.074)
Death Occurred -2660
in: 1998 (0.268)
1999 -10467
(1.056)
2000 -3180
(0.305)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

-32-

1"! If .

 

Table 4. Change in Value of Crop Income per Hectare

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death 430 -3471 4040 11969
(0.139) (0.839) (0.307) (0.752)
Sex of Deceased 8386 7091 9185
(1.417) (1.126) (1.349)
Age of Deceased -54.51 -176
(0.217) (0.526)
Education of -746 -742
Deceased (0.993) (0.983)
Deceased 4046
Member: (0.433)
Household Head
Spouse of Head 1593
(0.165)
Death Occurred -2900
in: 1998 (0.29)
1999 -12060
(1.21)
2000 -3338
(0.3 19)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

-83-

Table 5. Chage in Value of Total Household Income (Constant 1997 prices)

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death -45221 -21835 145424 163723
(1.355) (0.488) (1.024) (0.951)
Sex of Deceased -50272 -78976 -49547
(0.786) (1.160) (0.673)
Age of Deceased -3158 -4921
(1.164) (1.355)
Education of -5427 -4689
Deceased (0.668) (0.575)
Deceased 1 12940
Member: (1.1 19)
Household Head
Spouse of Head 21514
(0.206)
Death Occurred -468 13
in: 1998 (0.433)
1999 25 5 10
(0.237)
2000 582
(0.005)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

-34-

Table 6. Change in Value of Total Household Income

 

Regression 1 Regression 2 Regression 3 Regression4
Adult Death -37556 -17196 133008 145912
(1.209) (0.4 13) (1.005) (0.910)
Sex of Deceased -43767 -69539 -41736
(0.734) (1.097) (0.609)
Age of Deceased -2902 -4648
(1.148) (1.374)
Education of -4496 -3 839
Deceased (0.594) (0.505)
Deceased 10831 1
Member: (1 .152)
Household Head
Spouse of Head 25556
(0.263)
Death Occurred -32843
in: 1998 (0.326)
1999 25242
(0.252)
2000 1 1897
(0.1 13)

 

“Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database”, 1997 and 2000,
Tegemeo Institute, Nairobi, Kenya.

-35-

Bibliography

Adarns, Alayne M., Jindra Cekan and Rainer Sauerbom. "Towards a Conceptual Framework
of Household Coping: Reflections fi'om rural West Africa." Afiica, 68(2): 1998,
p.262-81.

Ainsworth, Martha, and Mead Over. "AIDS and Afiican Development." The World Bank
Research Observer, 9(2): July 1994, p.203-40.

Ainsworth, Martha, and Innocent Semali. "The Impact of Adult Deaths on Children’s
Health in Northwestern Tanzania." The World Bank Development Research
Group, Policy Research Working Paper 2266, Jan 2000.

Beaver, Margaret, Torn Awour, Mary Kiiru, et a1. Tegemeo Agricultural Marketing and
Policy Analysis Project, Household Survey Data Documentation, Michigan State
University, 2000, p.1-14.

Beegle, Kathleen G. "The Impact of Adult Mortality in Agricultural Households: Evidence
from Rural Tanzania." Dissertation, Michigan State University, 1997.

Bollinger, Lori, John Stover, and David Nalo. "The Economic Impact of AIDS." The
POLICY Project, USAID, 1999.

Bonnel, R. "HIV/AIDS and economic growth: A Global Perspective." The South Afiican
Journal of Economics. Special Edition, vol. 68:5, Dec 2000.

Brass, William and Carole Jolly, eds. Population dynamics of Kenya. National Research
Council. National Academy Press, Washington, DC. 1993.

Committee on World Food Security. "The Impact of HIV/AIDS on Food Security. " Rome,
June 2000. http://www.fao.org/docrep/meeting/003/y0310E.htm.

du Guemy, Jaques. "AIDS and Agriculture in Africa: Can Agricultural Policy Make a
Difference?" FAO, 2001.

Development News. "Agriculture, New AIDS Victim?" World Bank Development News,
Oct. 23, 2001.

Famine Early Warning Systems Network. "HIV/AIDS and Food Insecurity: Breaking the
Vicious Cycle." Greater Horn of Afiica Food Security Update, April 2000.

FAO. "I-HV/AIDS: a threat to sustainable agriculture and rural development". FAQ News and
Highlights, June 22, 2000.

-86-

 

 

Economic Digest 2000. Central Bureau of Statistics, Government of Kenya, 2000.

Mtika, Mike Matharnbo. "Social and Cultural Relations in Economic Action: the
embeddedness of food security in rural Malawi amidst the AIDS epidemic."
Environment and Planning A, 32: 2000 p. 345-360.

Lundberg, Mattias, and Mead Over. "Transfers and Household Welfare in Kagera." World
Bank, draft, June 2000.

McPherson, Malcolm F. "The Impact on Economic Growth in Afiica of Rising Costs and
Labor Productivity Losses Associated With HIV /AIDS." John F. Kennedy School of
Harvard University, September, 2000.

Nampanya-Serpell, Namposya. "Social and Economic Risk Factors for HIV /AIDS-Affected
Families in Zambia." Presented at AIDS and Economics Symposium, Durban,
IAEN network July 2000.

Baltazar, GM, J. Stover, TM okeyo, BON hagembe, R Mutenri, eds. "AIDS in Kenya:
Background, Projections, Impact and Interventions". NASCOP (National AIDS and
STDs Control Programme), Nairobi, Kenya, fifth edition, 1999.

Olshansky, SJ, B. Carries, RG Rogers, et a1. "Infectious Diseases: New and Ancient Threats
to Infectious Diseases." Population Bulletin, 52(2): July 1997, p. 1-34.

Over, Mead, et al. "The Consequences of Adult Ill-Health", in The Health of Adults in the
Developing World, eds. Feachem, Richard G. A., Christopher J. L. Murray, Margare
Phillips, Mead Over, Tord Kjellstrom..Oxford University Press and World Bank,
1992.

Rugalema, Gabriel. AHIV/AIDS and the Commercial Agricultural Sector of Kenya: Impact,
Vulnerability, Susceptibility and Coping Strategies. Institute of Social Studies, The
Hague, Netherlands, 1999.

Rugalema, Gabriel. "Coping or Struggling? A Journey into the Impact of HIV/AIDS in
Southern Afiica." Review of Afiican Political Economy, 86: 2000, p.537-45.

Rural Planning Department, Office of the Vice President and Ministry of Planning and
National Development, Kenyan District Development Plans. 1993-96.

Staatz, John M. "The Application of Social Capital in Understanding Economic
Development." MSU Agricultural Economics Staff Paper no. 98-17, 1998, p.5-11.

Stillwaggon, Eileen. "HIV Transmission in Latin America: Comparison with Africa and

Policy Implications." The South Afiican Journal of Econorrrics. Special Edition,
vol. 68:5, Dec 2000.

-87-

Stillwaggon, Eileen. "AIDS and Poverty in Africa." The Natiop, May, 21 2001.

"Tegemeo/MSU Agricultural Monitoring and Policy Analysis Household Survey database",
1997 and 2000, Tegemeo Institute, Nairobi, Kenya.

The World Bank. Intensifip'ng Action Against HIV/AIDS in Afiica: Responding to a
Development Crisis. 2000.

The World Bank. Confronting AIDS: Public Priorities in a Global Epidemic. 1997.

Tibaijuka, Anna Kajumulo. "AIDS and Economic Welfare in Peasant Agriculture: Case
Studies from Kagabiro Village, Kagera Region, Tanzania." World Development,
vol.25:6, June, 1997.

UNAIDS. A Review of Household and Community Responses to the HIV /AIDS Epidemic
in the Rural Areas of Sub-Saharan Afiica. Geneva, Switzerland, 1999.

University of Nairobi, Medical Library. "AIDS in Kenya." Sessional Paper No.4, 1997.

White, Joanna, and Elizabeth Robinson. "HIV/AIDS and Rural Livelihoods in Sub-Saharan
Africa: A Desk Review." Natural Resources Institute, University of Greenwich,
September, 1999.

-33-

t

II IIII IIII IIIII II III II IIIIIII IIIII I

 

02448 0679