Family meals and why they are worth the effort. What have we learned from Project EAT?

FAMILY HEALTH AND
WEALTH STUDY
INSIGHTS ON WEALTH MEASUREMENT
AND CHANGE
Februar y 2 0 , 2 01 3
STUDY AIMS
To assess the effect of childbearing patterns
on family health and wealth outcomes
Number and timing of births
Role of contraception
Family wealth and health outcomes
Household income, employment
Child schooling, nutrition
Maternal health
To assess using a longitudinal design
FHWS SITE LEADS/DATA COORDINATORS
Addis Ababa University
Assefa Seme
Meselech Roro
Obafemi Awolowo University
Peter Ogunjuyigbe
Abimbola Phillips
Assiut University
Omaima El Gibaly
Ghada Al- Attar
University of Ibadan
Michael Okunlola
Imran Morhason-Bello
Nathanael Afolabi
Kwame Nkrumah University of
Science and Technology
Easmon Otupiri
Denis Yar
Makerere University
Fred Makumbi
Vivian Zalwango
University of Malawi
Frank Taulo
Eddie Malunga
Wanangwa Chimwaza
FHWS ACKNOWLEDGEMENTS (US)
 Andreea Creanga
 Alain Koffi
 Funmi OlaOlorun
 Nadia Diamond Smith
 Qingfeng Li
 Adel Takruri
 Linnea Zimmerman
 Timothee Fruhauf
 And the rest of the
FHWS team
 Saifuddin Ahmed
 Michelle Hindin
 Stan Becker
 David Bishai
 Julia Driessen
 William Pan
STUDY DESIGN
Three rounds of observation
Probability sample of families in peri-urban
area
 Wife of childbearing age (15-49 years)
 Husband of childbearing age (20-54 years)
 GPS mapping of area (waypoints, households)
Data collection began by Ghana site January
2010
Round 2 approximately 2 years later
MEASUREMENT STRUCTURE AND CONTENT
 Household roster on occupants and their characteristics, GPS
 Focal woman questionnaire





Background characteristics
Childbearing history, fertility preferences and contraceptive calendar
Child schooling (5 to 24 years) and health history (births in <5 years)
Marital relationship quality, decision-making autonomy
Self-reported health
 Focal man questionnaire




Background characteristics
Parity, fertility preferences and contraceptive use
Marital relationship quality, decision-making autonomy
Adult morbidity and self-reported health
 Wealth module
 Housing construction quality, asset ownership, expenditures in <1 year
 Physical assessment
 Height, weight of household members
 Blood pressure, pulse
 Anemia (Ghana, Uganda)
PERI-URBAN FHWS SITES
Country
Site
Egypt
Ethiopia
Ghana
Malawi
Nigeria
Nigeria
Uganda
Total
Waldeya
Sebeta
Asawasa
Lunzu
Ipetumodu
Akinyele
Wakiso
Sample size
548
998
800
605
787
502
505
4745
SITES VARIED IN ‘PERI-URBAN-NESS’
TRAINING IS EVERY THING
FHWS Round 2 training for Sebeta site
LEARNING TO TAKE BLOOD PRESSURE,
MAPPING AND COMMUNIT Y SENSITIZATION
July 2011 workshop
Blantyre, Malawi
PRECISION AND ENTHUSIASM
KNUST FHWS Team and some of equipment field
staff transported during interviews
PROGRESS TO DATE
 Analysis workshops in July 2011 and July 2012
 Data sharing and authorship agreements
 Gates Institute role is facilitating comparative analyses on pre-defined
set of topics
 Panel at International FP Conference 2011, Dakar
 Each site has autonomy to share data with analysts within and outside
 Ghana: Two dissertations
 Ethiopia: One dissertation
 One year spent on data cleaning and linking rounds
 Follow-up rates and who is missed
 Analyses under way





Comparative description of 4745 families’ health and wealth
Childbearing patterns and child schooling and nutrition
Childbearing and family wealth
Couple concordance in fertility preferences and contraceptive use
Parity and self-rated health (and gender differences)
Parity Composition among Married Women Aged 15‐44 in Six FHWS Sites
100%
90%
80%
70%
60%
6+
4‐5
50%
2‐3
40%
0‐1
30%
20%
10%
0%
Ibadan
Ife
Kumasi
Lunzu
Wakiso
Sebeta
Modern Contraceptive Prevalence among Married Women Ages 1544 by Age Group across Six Africa-based FHWS Sites
% using modern contraceptive method
80.0
70.0
60.0
50.0
Ibadan
Ife
Kumasi
Lunzu
Wakiso
Sebeta
40.0
30.0
20.0
10.0
0.0
15-19
20-24
25-29
30-34
Age group
35-39
40-44
Modern Contraceptive Prevalence among Married Women Ages 15‐
44 by Parity across Six Africa‐based FHWS Sites
% using modern contraceptive method
80.0
70.0
60.0
50.0
Ibadan
Ife
40.0
Kumasi
Lunzu
30.0
Wakiso
Sebeta
20.0
10.0
0.0
0‐1
2‐3
4‐5
Parity group
6+
CAPTURING HOUSEHOLD TRANSITORY
WEALTH THROUGH AN INDEX ON
EXPENDITURES AND NON-DURABLES
J. Driessen, P. Ogunjuyigbe, A. Phillips, Q. Li, FHWS Study Team, A. Fatusi, A. Tsui
ANALYSIS OF PERMANENT AND
TRANSITORY WEALTH MEASURES
Common use of wealth quintiles from assets
assessed in surveys (EDHS)
Wealth measure can be broken down into
Permanent wealth (house, housing quality,
vehicle, ownership of durable goods)
 Transitory wealth (expenditures on entertainment,
eating out, other consumption reflective of ‘middle
class’ lifestyle)
 Such data are challenging to collect
 Proxied with asset ownership of durables and nondurables, expenditures, income, household quality
RATIONALE FOR INDEX CONSTRUCTION
 Address overlapping measurement of wealth
 Deconstruct household wealth into permanent and
transitory components
 Create a summative index
 Selection of index items
 Weights for each item
 Dichotomous versus continuous measures
 DHS wealth quintiles based on PCA with dichotomous
measures
 Used principal components analysis
OUTCOMES OF INTEREST
 Self-rated wealth
 9-step ladder of perceived relative economic status
 Satisfaction with current income
 4-step rating scale
 Aspirational wealth
 5-step rating scale of relative well-being in one year
PERCENT DISTRIBUTION FOR
SATISFACTION WITH CURRENT INCOME
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Fully satisfied
Rather satisfied
Less than satisfied
Not at all satisfied
ANALY TIC APPROACH
 Regress self-reported economic wellbeing measures
on
 Permanent wealth index (Fixed Asset Index)
 Transitory wealth index (Middle Class Index)
 Covariates
 Male years of schooling
 Number of persons in HH under age 5
 Number of persons in HH age 5-14
PCA RESULTS FOR FIXED ASSETS
Sites
Eigenvalue of component 1
% variance explained by component 1
Cronbach's alpha
Range of predicted score (min max)
Ethiopia
4.89
10.4
0.774
(-3.92
9.78)
Ghana
3.27
7.4
0.638
(-4.22
6.14)
% variance explained by component 1
with all fixed asset and middle class
index items
9.0
5.9
Malawi Nigeria/Ife
4.67
2.87
14.1
6.2
0.758
0.575
(-2.67
(-5.13
17.44)
6.12)
12.9
6.1
Nigeria/
Ibadan
3.40
8.5
0.604
(-4.10
11.06)
Uganda
4.28
10.4
0.710
(-3.20
9.99)
8.2
13.7
EXAMPLE OF FACTOR LOADINGS FOR
FIXED ASSET INDEX
Nigeria/ Nigeria/
Ethiopia Ghana Malawi
Ife Ibadan Uganda
Furnishings
Has bed
Has table
Has chair
Has dresser
Has refrigerator
Has landline telephone
Has motorcycle
Has bicycle
Has car/truck
Has horse cart
Has generator
0.12
0.08
0.08
0.26
0.29
0.28
0.01
0.09
0.20
-0.01
0.10
0.00
-0.03
0.01
0.13
0.24
0.04
0.03
0.01
0.12
0.00
0.03
0.21
0.21
0.19
0.17
0.30
0.15
0.06
0.05
0.25
‐‐
0.07
0.21
0.24
0.23
0.18
0.30
0.04
0.04
0.02
0.26
0.03
0.29
0.24
0.27
0.26
0.20
0.25
0.14
0.03
0.07
0.26
‐‐
0.27
0.14
0.21
0.19
0.23
0.30
0.14
0.06
0.13
0.29
0.15
0.20
PCA RESULTS FOR MIDDLE CLASS INDEX
Sites
Eigenvalue
% variance explained by component 1
Cronbach's alpha
Range of predicted score (min max)
Ethiopia
3.79
11.2
0.671
(-2.67
12.64)
Ghana
3.60
10.6
0.660
(-1.57
18.15)
Malawi
4.83
15.1
0.755
(-2.12
11.82)
Nigeria/ Nigeria/
Ife Ibadan Uganda
3.20
3.30
7.67
9.4
9.7
22.6
0.677
0.668 0.522
(-4.09
(-2.70 (-0.93
7.77) 14.61) 23.65)
EXAMPLE OF FACTOR LOADINGS FOR
MIDDLE CLASS INDEX
Sites
Consumption/expenditure behaviors
Spent >$2.5 eating out in last 7 days
Spent >$10 in last month on clothes/shoes
Spent >$10 in last month on daily
household items
Spent >$5 in last month on medicines
Spent >$10 in last month on books,
newspapers, school supplies and
entertainment
Spent >$5 in last month on other products
and services
Spent >$20 in last month on child care
Spent >$15 in last 7 days on food (less
amount spent eating out)
Spent >$10 in last month on utilities
Paid any amount for taxes last year
Household has no debt
Household has lent any amount to others
Household currently has savings
Nigeria/ Nigeria/
Ife Ibadan Uganda
Ethiopia
Ghana
Malawi
0.19
0.07
0.30
0.29
0.14
0.14
0.21
0.22
0.10
0.16
-0.02
-0.01
0.19
0.07
0.30
0.30
0.20
0.14
0.30
0.23
0.15
0.09
-0.01
0.01
0.18
0.22
0.23
0.22
0.16
0.00
0.12
0.13
0.24
0.26
0.05
0.24
0.22
0.20
0.13
0.10
-0.01
0.00
0.23
0.28
0.23
0.01
0.05
0.18
-0.25
-0.01
0.26
-0.16
0.11
0.08
0.20
0.25
0.20
0.00
0.07
0.22
0.07
0.20
0.17
-0.06
0.11
0.14
0.13
0.24
0.15
-0.04
0.14
0.18
0.00
0.01
-0.02
0.00
-0.01
-0.01
Red frame indicates statistical significance at 5% level.
Adjusted for male education, presence of children and youth in household
Red frame indicates statistical significance at 5% level.
Adjusted for male education, presence of children and youth in household
Red frame indicates statistical significance at 5% level.
Adjusted for male education, presence of children and youth in household
WHAT WE’VE LEARNED ABOUT WEALTH
MEASUREMENT
 PCA can be applied to other non-asset variables and
reduce reliance on reported expenditure data
 Constructed Middle Class Index reflective of consumption
and short-term well-being
 MCI performs similarly to Fixed Asset Index in predicting
self-reported wealth outcomes
 Middle class index sensitive to
 Expenditure ‘shocks’ (e.g., unanticipated health expenses)
 Health expenditure shock likely associated with having
sick children
 Presence of children
 Children may drive expenses captured in transitory
wealth score
0
.05
Density
.1
.15
.2
.25
THE DISTRIBUTION OF THE PERMANENT WEALTH
SCORES IN ROUND 1 AND 2 (ETHIOPIA)
-5
0
5
Fixed Assets Scores
Round 1
Round 2
10
0
.05
.1
Density
.15
.2
.25
THE DISTRIBUTION OF THE TRANSITORY WEALTH
SCORES IN ROUND 1 AND 2 (ETHIOPIA)
-5
0
5
MCI Scores
Round 1
10
Round 2
15
FACTORS ASSOCIATED WITH COUPLE
LOSS-TO-FOLLOW UP (ETHIOPIA)
Factor
Family income (Round 1/log)
Length of residence (Husband)
Length of residence (Wife)
Borrowed money last year for health expenses
Own house
Duration of marriage (Wife report)
Regret marrying spouse (Husband)
Regret marrying spouse (Wife)
Husband has other wives (Wife report)
Husband has other wives (Husband report)
Adj OR
p level
1.011
0.997
0.988
0.609
0.619
0.953
1.470
1.358
1.265
0.474
0.011
0.687
0.199
0.051
0.010
0.001
0.164
0.164
0.543
0.211
*Model also controls for occupation type of husband and wives (ns)
n=950 couples, weighted for loss to follow up
ROC (RECEIVING OPERATING
CHARACTERISTICS) CURVE TO ASSESS
PROPENSIT Y SCORE
.75
1
Closeness of curve to diagonal line
is favorable to constructed
propensity score model
0
Density
2
3
.25
4
Sensitivity
.5
Distribution of propensity scores of
those missed and relocated
are similar
(Ethiopia results only)
.25
.5
1 - Specificity
.75
1
1
0
0
Area under curve = 0.6743 se(area) = 0.0196
0
.2
.4
Propensity Score
Captured in 2nd round
Missed in 2nd round
.6
.8
FACTORS AFFECTING ROUND 2
TRANSITORY WEALTH SCORE
Factor
Middle class score (Round 1)
Fixed asset score (Round 1)
Husband's years of education (Round 1)
Borrowed money last year for health
expenses
Number of children < age 7
1
2
n=693 couples, weighted for loss to follow up
Reg Coeff
p level
0.777
0.088
0.020
0.000
0.000
0.007
-0.126
0.122
0.128
0.088
0.061
0.365
WHAT WE’VE LEARNED ABOUT CHANGE
THROUGH FHWS
 Critical importance of training and supervision
 It’s not worth doing, if it’s not done well
 Standardized data-entry formats
 Importance of longitudinal study design
 Under-estimated loss-to-follow up which impacts Round 2 sample
power
 Challenges with relocating couples in peri-urban areas
 Couple follow-up rates are not surprisingly lower than individual
follow-up rates
 Household loss due to logical events (marital disruption, migration,
death)
 Ability to decompose overall wealth into permanent and
transitory components
 This type of study is rare in the African setting.
WHAT WE HAVEN’T LEARNED AS YET
Whether childbearing patterns consistently
influence family health and wealth outcomes
THANK YOU AND MANY THANKS
TO THE PRODIGIOUS EFFORTS OF THE
EXTENDED FAMILY OF RESEARCHERS AND FHWS STUDY
PARTICIPANTS AND SUPPORT FROM
THE BILL & MELINDA GATES FOUNDATION
THROUGH THE GATES INSTITUTE.