Title: Calibrating orphanhood: the number of orphans according to recent censuses and health surveys alread
1Calibrating orphanhood the number of orphans
according to recent censuses and health surveys
already exceed UNAIDS estimates for 2010 for
Kenya and Benin and 4/5th for South Africa
Robert McCAAFélicien Donat Edgar T.
ACCROMBESSYKhassoum DIALLOcontact
rmccaa_at_umn.edu
2The 2002 census of Benin reveals as many orphans
as predicted for 2010
UNAIDS 2004estimatesfor 199019952000200
3 and2010
The UNAIDS estimates show a progressive rise
1990-2010Note the 1990 census of Benin did not
contain a question on orphanhood.
3Kenya orphanhood statistics UNAIDS estimates
1990-2010compared with 1989 and 1999 census
dataage 0-17 orphans of any condition
Kenya 1999 censused orphans 2010 est
1999 census
UNAIDS 2004estimatesfor 199019952000200
3 and2010
1989 census
Sources Kenya CBS-Kenya census sample from
www.ipums.org/internationalUNAIDS/UNICEF/USAID
Children on the Brink 200432. Note Results
depend upon interpretation of Do not know and No
reply. We use UNAIDS rule DNK or NR dead .
4What is original about this paper
Compare orphanhood estimates from Demographic and
Health Surveys DHS with census microdata CMD
to find that the two sources are in close
agreement by age pattern and type maternal
paternal and double Reveal that CMD figures are
higher than UNAIDS estimates of orphanhood in
some cases higher than the 2010 estimates.
Show that census microdata can be used but are
not used by UNAIDS to omitted for lack of
time Study the condition of orphans in household
s Compare trends over time between and within co
untries Conclude that the 2010 round of censuses
may provide important benchmarks for the UNAIDS
estimatesif the orphanhood relation to
hourseholder and associated questions are
retained and if researchers analyze them.
5Outline the number of orphans in recent
censuses and health surveys already exceed
UNAIDS estimates for 2010 for Kenya and Benin
4/5th for South Africa
The orphan problem numbers and consequences
Statistics and sources UNAIDS vs. microdata
UNAIDS estimated projectionsbased on
demographic-epidemiological models
Surveys Demographic Health
Censuses and census microdata example of S.
Africa Is your mother/father alive?Why not
use these data to estimate/study orphans?
Insights from census microdata
More orphans now than projected for 2010
Yet the extended family still shelters orphans
Policy implications aid/assist all children
6The orphan problem caused by AIDS how many and
how will they be cared for
UNAIDS 1990-2010 projects a tripling of the
number of orphans even with reduced fertility
due to a demographic transformation b AIDS as
well as c increased mortality of children
Given the low life expectancy in many African
countries the extended family particularly the
grandmother has long cared for orphans.
With the surge in the number of orphans and life
expectancy contracting will the extended family
continue to shelter orphans?
7Deconstructing orphan
In African culture we do not make the same
distinctions among relations practised by whites.
We have no half-brothers or half-sisters. My
mothers sister is my mother my uncles son is
my brother my brothers child is my son my
daughter. --Nelson Mandela 1994 What is an or
phan? --a social construct on a biological base
Mandela or models kin terms and relations--the
adoption effect Sociological may include foste
r/step/etc. vs. bio/demographic
UNAIDS definition child least 1 parent is dead What are the implications?
Demography mortality requires biological defin
ition Brass 1971 Policy to promote child welfa
re is a sociological definition ok?
Sources fuzziness of the census may be an
advantage for policy making although a
disadvantage for estimating mortality
8Personal anecdotes
Low number of orphans due to AIDS 2003
23-48000 5-10
93 Sources on orphans
UNAIDS method 2004209 demographic-epidemiolog
ical model developed by UNAIDS Reference Group on
HIV/AIDS Estimates Modeling and Projections
with plausibility bounds Fertility mortality AI
DS mortality Estimate maternal orphans followed
by estimates of paternal orphans using male
fertility patterns for ages 0-17 years
Concordance with DH Surveys Grassly et al
2005373 DHS the gold standard Periodic f
inely tuned comprehensive instruments skilled
interviewers many countries ages 0-14 only.
Census National in scope snapshot of basic
demographic and social characteristics of
individuals all ages families/households.
Poorly trained interviewers simple questions
nothing on AIDS mortality estimates are
contentious Microdata are difficult to obtain A
CAP few 2000 round censuses IPUMS 1 African
country census agencies
10Benin a West African example of low prevalence
of AIDS yet number of orphans in 2002 already
exceeds UNAIDS 2010 estimate
Low number of orphans due to AIDS 2003
23-48000 5-10 HIV positive in capital city
60 of sex workers only 2 of young pregnant
women. Yet the total number of orphans in 2002
385000 exceeds the number projected for 2010
370000
11Kenya 1999 census questionnaire
12Kenya 1989 census instructions to enumerators
regarding orphanhood
Kenya 1989 Columns P17 and P18 -
Orphanhood 117 Is this persons father/mother a
live ? 118 Code 1 or 2 in respect of the person
s biological father and mother. Foster parents
or other relatives who may have adopted the
person should not be considered as the father or
mother of the person emphasis added.
119 In some cases childs father may not be
married or living with the mother. In this case
the mother might report that she does not know
whether the father of her child is alive or dead.
In this case code 3 for Not-Known.
Source www.ipums.org/international/enumforms/ken
/ken1989_enuminstruct.pdf frame 12.
Note the instructions clearly emphasize
biological parentage but did
enumerators pay attention? Better would be to i
nclude biological on questionnaire
131999 instructions to census enumerators regarding
orphanhood
Kenya 1989 Columns P17 and P18 -
Orphanhood 117 Is this persons father/mother a
live ? 118 Code 1 or 2 in respect of the person
s biological father and mother. Foster parents
or other relatives who may have adopted the
person should not be considered as the father or
mother of the person emphasis added.
119 In some cases childs father may not be
married or living with the mother. In this case
the mother might report that she does not know
whether the father of her child is alive or dead.
In this case code 3 for Not-Known.
Source www.ipums.org/international/enumforms/ken
/ken1989_enuminstruct.pdf frame 12.
Kenya 1999 Columns P20-21 Orphanhood
79. Is this persons father/mother alive?
a Tick the box under the appropriate column in
respect of the survival status of the
respondents biological father and mother. Note
that at times destitute children are brought up
or adopted at a very young age by relatives. Such
foster parents should not be considered as the
biological parents of the respondent. Please
always probe to establish the reality of the
situation emphasis added. b In some cases a
childs father/mother may not be married or
living with the mother/father. In this case the
mother/father might report that she/he does not
know whether the father/mother of her child is
alive or dead. In this case mark an X in the
box for dont know. You must always probe to
ensure you obtain the most satisfactory answer.
Source www.ipums.org/international/enumforms/ken
/ken1999_enuminstruct.pdf frame 31.
143.1
2.2
2.5
- Sources Demographic and Health Survey DHS
Kenya 1998 Table 2.4 1999 census microdatata
imputed come from a custom non-circulating
edited sample supplied by the Central Bureau of
Statistics of Kenya 1999 and 1989 are from
harmonized extracts from https//www.ipums.org/int
ernational
15- Sources UNAIDS Children on the Brink 2002 p.
16 22 and 2004 p. 26 30 1999 census
microdatata with imputed come from a custom
non-circulating edited sample supplied by the
Central Bureau of Statistics of Kenya 1989 and
1999 figures are from a harmonized extract
obtained from www.ipums.org/international . In
both instances do not knows are recoded to
dead that is orphanhood is maximized. UNAIDS
convention is total orphans paternal
maternal double. We have converted such
figures thus
total orphans paternal
maternal double.
16- Sources UNAIDS Children on the Brink 2002 p.
16 22 and 2004 p. 26 30 1999 census
microdatata with imputed come from a custom
non-circulating edited sample supplied by the
Central Bureau of Statistics of Kenya 1989 and
1999 figures are from a harmonized extract
obtained from www.ipums.org/international . In
both instances do not knows are recoded to
dead that is orphanhood is maximized. UNAIDS
convention is total orphans paternal
maternal double. We have converted such
figures thus
total orphans paternal
maternal double.
17Recall Kenya orphanhood statistics UNAIDS
estimates 1990-2010compared with 1989 and 1999
census dataage 0-17 orphans of any condition
Kenya 1999 censused orphans 2010 est
1999 census
UNAIDS 2004estimatesfor 199019952000200
3 and2010
1989 census
Sources Kenya CBS-Kenya census sample from
www.ipums.org/internationalUNAIDS/UNICEF/USAID
Children on the Brink 200432. Note Results
depend upon interpretation of Do not know and No
reply. We use UNAIDS rule DNK or NR dead .
18South Africa orphanhood UNAIDS estimates
1990-2010compared with 1996 and 2001 census
microdataage 0-17 orphans of any condition
2001 census is four-fifths of estimate for 2010
UNAIDS 2004estimatesfor 199019952000200
3 and2010
2001 census
Anomaly? Or history?
1996 census
19South African census microdata1996 and 2001
compared
1996 maternal orphanhood data are good Bah
199936 -- does not examine paternal.
For 1996 2001 low rates of Do Not Know and
No Reply Increase in total orphans is less than
expected but both are higher than UNAIDS
estimates millions 2.2 96 vs. 1.5 for 95
and 2.5 01 vs. 2.2 for 03
High concordance with DHS see nexts
similar by age pattern 0 1 17 by type mat
ernal paternal double
20Are census microdata useful/reliable?Compare
orphanhood microdata by age of orphan and type
paternal maternal double for South Africa
A gold standard DHS Survey 1998
21Compare DHS 98 with Census Microdata
1996Amazing agreement Census figures are more
regular
1996 census microdata closely track the 1998 DHS
22Add Census Microdata for 2001Amazing agreement
2001 rates are indeed higher
2001 census microdata are predictably higher
23Why arent census microdata being used to
calibrate UNAIDS estimates
For Africa census microdata are difficult to
obtain No recent census Angola Cameroon DR Co
ngo Liberia Nigeria etc. Dissemination of 200
0 round microdata limited to 3 countries so far
Kenya Mauritius and South Africa Benin made
available to co-author. ACAP Penn and IPUMS-In
terntational Minnesota seek to provide access
to researchers Perhaps too researchers consid
er census data as less reliable
Or perhaps they are not familiar with them.
24Conclusions
For the 2000 round of censuses
Microdata should be used as they become
available Commercial www.ipums.org/internationa
l offers Kenyan census microdata plus other
countries to researchers w/o cost South Africa
coming soon. For the 2010 round of censusesthe
microdata will be a valuable benchmark
If the censuses are conducted If they include que
stions on orphanhood If researchers use the censu
s microdata Policy implications measure manage
Paris21 There will be more orphans than UNAIDS
estimates Currently families shelter orphans a
t some cost and not entirely satisfactorily but
can they in 2010? Thank you