Measuring the Impact of Young Adult
Mortality on the Wellbeing of Older
Persons in KwaZulu-Natal, South Africa
Marjorie Opuni-Akuamoa
Advisor: Dr David Bishai
Department of Population, Family and Reproductive Health
Johns Hopkins Bloomberg School of Public Health
First Annual Research Conference,
Population Impacts on Economic Development
London, 1-3 November, 2006
Presentation Outline
• Study Background
• Specific Aims
• Conceptual Framework
• Study Data
• Research Methods
• Strengths and Limitations
Increase in Young Adult Mortality
• Increase in young adult mortality in
Southern Africa over last 20 years
• In KwaZulu-Natal, South Africa
– The probability of dying between 15 and 60
years was 58% for women and 75% for
men in 2000 (Hosegood et al., 2004)
• In Zimbabwe
– The probability of dying between 15 and 60
years was 50% for women and 65% for
men in 1997 compared to 33% for both in
1985 (Feeney, 2001)
Log Odds of Dying, 2001
Northern KwaZulu-Natal, South Africa
Case and Ardington, 2005
Demographic Consequences of
Increase in Young Adult Mortality
• Orphans
– Millions of children (under the age of 18)
have lost one or both parents
– In sub-Saharan Africa the number of
orphans has increased by one third since
1990 (UNAIDS, UNICEF, USAID, 2004)
– 13% of South African children were orphans
in 2003 (projected to increase to 19% by
2010) (UNAIDS, UNICEF, USAID, 2004)
Demographic Consequences of
Increase in Young Adult Mortality
• Bereaved Parents
– Many older adults (50 years and above)
are living to see their adult children die
rather than the reverse
– South African women aged 60 years and
above with no surviving children projected
to increase from 10% to 20% between
1995 and 2010 (Merli and Palloni, 2004)
Young Adult Mortality
and Household Members
• Ill health and mortality among adults of
reproductive age can affect:
– Household composition
– Household labor allocations
– Household income, assets and
savings
– Household consumption patterns
– Health and wellbeing of household
members
Over et al., 1992
Young Adult Mortality
and Persons Aged 50+ Years (1)
• A lot of media and advocacy attention to
grandparents caring for orphans
• Few quantitative studies on the
association between young adult
mortality and the health and wellbeing of
older adults (Knodel et al., 2001, 2002, 2003; Ainsworth and Dayton,
2003, 2004; Hosegood and Timaeus, 2005)
Young Adult Mortality
and Persons Aged 50+ Years (2)
• Older persons rely on their children and
grandchildren for money, food and other
support
• Illness and mortality of young adults can
result in reductions in support while
expanding care-giving and financial
burdens of older adults
Rationale for Study
• “Without a better research foundation,
developing appropriate programs and
policies to address the special needs of
older persons or to use their potential as
human resources in addressing the
consequences of the dramatic
increases in young adult mortality will
be difficult” (Knodel, Watkins, Van Landingham, 2003)
Specific Aims (1)
• To assess whether the occurrence
of a young adult death affects the
probabilities of not working, working
part time and working full time for
non-pensionable older women
(aged 50-59 years) and non-
pensionable older men (aged 50-64
years).
Specific Aims (2)
• To assess if the occurrence of a
young adult death is associated with
the value of assets of households
where persons aged 50 years and
above live.
Specific Aims (3)
• To examine if the occurrence of a
young adult death is affects the
self-reported health status of
persons aged 50 years and above.
Conceptual Framework
Distal Determinants Proximate Determinants
Health Food intake by older
older person
person
time t-1
Community Medical care used
endowment by older person
Household Time used to
endowment produce health Health
by older person older
person
time t
Level of education Individual
of older person endowment of older
person
Health Genetic endowment
young of older person
person
time t-1
Economic Model (1)
Where:
Hi= Health of household member i
Xi= Consumption of household member i
TiL= Leisure time of household member i
n= Number of individuals in household
Max U U ( H i , X i , Ti ), i 1,...n
L
Ij= Non wage income
wk= Vector of market wages in community k
s.t. Ti= Total time household member i has to
allocate to work, leisure and health
n n n Pk= Vector of prices in community k
I j Ti wk X i pk Ti wk
L Ni= Food intake of household member i
Mi= Medical care used by household
i 1 i 1 i 1 member i
Ti H= Time used by household member i on
and health
Ei= Level of education of household
H i H ( N i , M i , Ti , Ei , Ci ,i , C j , Ck )
H member i
ηi= genetic endowment of household
member i
Ci=vector of other individual endowments
Cj=vector of household endowments
Ck=vector of community endowments
Behrman and Deolalikar, 1988 adapted by Ainsworth and Dayton, 2003
Economic Model (2)
• Maximizing household utility subject to
the budget constraint and the health
production function, the following
reduced form equation is obtained:
Hi H ( Ei , Ci ,i , C j , I j , Ck , pk , wk )
*
Study Data
• KwaZulu-Natal Income Dynamics
Survey (KIDS)
– University of KwaZulu-Natal (UKZN), University of
Wisconsin-Madison, International Food Policy
Research Institute (IFPRI), London School of
Hygiene & Tropical Medicine (LSHTM) and
Norwegian Institute of Urban and Regional
Studies (NIBR)
– Household surveys - Living Standard
Measurement Survey - 1993, 1998, 2004
– 1,354 households in 1993; 1,132 households in
1998; 841 households in 2004
– Community surveys – 1993, 1998, (2004)
KwaZulu-Natal, South Africa
• Population: 9.5
million (20% of South
African population)
• Ethnic composition:
85% African; 12%
Indian
• GNP per capita:
9,713 Rand (approx.
1,300 USD)
• ANC HIV prevalence:
41%
Methods: Aim 1 & Aim 3
• Data: KIDS, 2004
• Outcomes: Employment status (Aim 1) and
self-reported health status (Aim 3)
• Method: Ordered probit
• Covariates include: for older persons –
gender, age, marital status, education; values
of assets; unearned income; unemployment at
community level; measures of community
health and infrastructure; mortality of elderly in
households; occurrence of young adult death
– 1-2 years and 3-4 years before the survey
Methods: Aim 2
• Data: KIDS, 1998, 2004
• Outcome: Value of assets
• Method: Fixed effects
• Covariates include: for older persons –
gender, age, marital status, education;
unearned income; unemployment at
community level; measures of community
health and infrastructure; mortality of elderly in
households; occurrence of young adult death
– 1-2 years and 3-4 years before the survey
Instrumental Variable Estimation
• Research on impact of young adult
illness and mortality plagued by three
methodological concerns: unobserved
heterogeneity, selection and reverse
causality
• All of these can cause endogeneity
• Solution:
– Instrumental variable estimation (Lundberg
et al., 2000)
– Propensity score matching
Strengths
• Detailed socioeconomic information
for households
• Population based data and not
convenience sample
• Systematic analysis attempt to
control for endogeneity of young
adult mortality
Limitations
• Limited health outcomes for older
persons
• Self-report - measurement error
• Limited cause of death information
• Measure impact of young adult
mortality within household only
Thank you!