An Own-Children Maternal Orphanhood Method For ...An Own-Children Maternal Orphanhood Method for...

Post on 07-Apr-2020

2 views 0 download

transcript

An Own-Children Maternal Orphanhood Method For Estimating

Fertility Rates from Census Microdata

Michel Garenne1 and Robert McCaa2

MPC Seminar Series January 22, 2016

1Institut de Recherche pour le Développement, UMR Résiliences, Paris Institut Pasteur, Epidémiologie des Maladies Emergentes, Paris Witwatersrand University, School of Public Health, Johannesburg 2Minnesota Population Center

April 5,

I P U M S 2 S 0 P 0 A 5 R I S

An Own-Children Maternal Orphanhood Method for Estimating Fertility Rates from Census Microdata • Objectives and Motivation

• Bob – promote need for and use of IPUMS census microdata • Michel - expand demographic tool kit using empirical data

• Empirical results, ZA: Authorities, OWCHMOM, OWCH • Methodology and Data:

• Method: Own Children, with mortality estimates from % maternal orphanhood, not model life tables • DATA: IPUMS Microdata only; use:

•  MOMLOC to match children to moms •  % maternal orphans to estimate mortality

•  Other examples: Burkina Faso à Zambia •  Conclusions, lessons learned

• MOMLOC is great! • It’s the mortality, stupid!

Bob’s job: Get data The pitch: Power of census microdata and value added by IPUMS Publications: Topics 1.  Microdata revolution - 2002

2.  Confidentiality – 2003

3.  Assortative mating – 2005

4.  Coherence – 2013-16

5.  Own children maternal mortality method?

Orphanhood questions: South Africa 2011 census

Almost all 2010 round African census samples have necessary variables. Exceptions: Egypt 2006, Ghana 2010, Nigeria 2010. An"X"indicatesthevariableisavailableinthatdataset.

Variable VariableLabel BF CM EG ET GH KE LR MW ML MZ NG ZA SS SD ZM200620052006200720102009200820082009200720102011200820082010

AGE Age(Singleyear) X X X X X X X X X X X X X X XCHBORN Childreneverborn X X . X X X X X X X . X X X XMORTMOT Mortalitystatusofmother X X . X . X X X X X X X X X X

MOMLOC Mother'slocaRoninhouse X X X X X X X X X X X X X X X

STEPMOM Probablestepmother X X X X X X X X X X X X X X XComputeOCHMOM?? Yes Yes No Yes No Yes Yes Yes Yes Yes No Yes Yes Yes Yes

Gold Standard: Moultrie & Timaeus, Pop Studies, 2003

Adjustments to 1996 census sample data:

1.  Weighted to compensate for undercount

2.  El Badry corrections for childlessness

3.  Discount stillbirths (estimated from ‘98 DHS)

Methods (not OWCH):

a.  Births last year (Census and DHS 1998)

b.  Reverse Survival (age structure + life tables)

0

12

34

56

TFR

1980 1990 2000 2010year

Moultrie & Timmaeus 2003 ochmom1996Sibanda & Zuberi 1999 ochmom2001Mostert et al. 1998 ochmom2011Udjo 1997 StatsSA 2015

Authoritative and Own Children Maternal Orphanhood Estimates From 1996, 2001, & 2011 MicrodataSouth Africa: Total Fertility Rates Compared

01

23

45

6TF

R

1980 1990 2000 2010year

Moultrie & Timmaeus 2003 ochmom1996Sibanda & Zuberi 1999 ochmom2001Mostert et al. 1998 ochmom2011Udjo 1997 StatsSA 2015

Authoritative and Own Children Maternal Orphanhood Estimates From 1996, 2001, & 2011 MicrodataSouth Africa: Total Fertility Rates Compared

OWCHMOM à higher than most. % maternal orphanhood suggests moderately lower life expectancy.

01

23

45

6TF

R

1980 1990 2000 2010year

Moultrie & Timmaeus 2003 owchmom1996Sibanda & Zuberi 1999 owchmom2001Mostert et al. 1998 owchmom2011Udjo 1997 StatsSA 2015Palamuleni 2013 (Rele method)

3 Own Children Maternal Orphanhood Series 1996, 2001 and 2011South Africa: Published Total Fertility Estimates Compared with OWCHMOM

Palamuleni’s trend 1996-2011 is an outlier

01

23

45

6TF

R

1980 1990 2000 2010year

owchmom1996

Moultrie & Timmaeus 2003

Reverse Survival (official StatsSA annual mortality estimates) vs. OWCHMOMSouth Africa: 2 Fertility Methods Compared

•  Was annual life expectancy ~5% (3 years) worse than official estimates?

•  If so, fertility estimates merge

•  Or is OWCHMOM (% maternal orphan) a bit high?

01

23

45

6TF

R

1980 1990 2000 2010year

owchmom1996Moultrie & Timmaeus 2003OWCH Princeton South StatsSA e0

Reverse Survival vs. OWCHMOM vs. OWCH South Princeton Models StatsSA e0 1981-1996South Africa: 3 Fertility Methods Compared

•  Add OWCH using StatsSA & Princeton South

•  Reverse Survival?

•  OWCH? •  OWCHMOM? •  ???

Rationale of Own Children Methods Reconstruct age specific fertility rates from a census inflating numbers of children using mortality rates derived from model life tables •  Age Specific Fertility Rates = Births / Women by age

and period •  Backward project person-years lived by women •  Backward project births from survivors •  Calculate ASFR and TFR for the 10 to 15 years before

the census

Framework

No birth ← Women → Died

↓ Living

elsewhere ← Births → Died

↓ Living at

home

Own Children Maternal Orphanhood Method - OWCHMOM

1.  Use only information available in census microdata

2.  Match mothers to co-resident children using MOMLOC (EASWESPOP is very good, MOMLOC is more accurate!)

3.  Backward project children ever-born directly (use spreadsheet, not EASWESPOP)

4.  Backward project women from orphanhood question (use spreadsheet, not EASWESPOP model life tables)

Data needed—3 computed by IPUMS Tabulator. The 4th (age by age_mom) would be cool

1.  AGE by SEX, age structure: number of women by single year of age: 12-64

2.  CEB, fertility: Mean number of children ever-born by AGE (single year), for women 12-64

3.  AGE by AGE_MOTHER: Age of children (0-49) by AGE of mother (12-64). Careful: MOMLOC where STEPMOM=0!

4.  MORTMOT, mortality (orphanhood): proportion of persons age 0-49 whose mother is alive (by single year of age)

Garenne OWCHMOM spreadsheet:

Calculations: 1) Women

Principle: start from cohorts of surviving women, and take into account their mortality since delivery Calculate mean age of children by age of mother = duration of exposure to mortality since birth of children Calculate the proportion of women who have died since delivery = proportion of children who are orphans (from orphanhood question, MORTMOT) Backward project person-years lived by women, by age and period

Calculations: 2) Children

Principle: start from children ever-born (and not from survivors) Calculate the distribution of surviving (own) children by age (year) and age of mother (year) Distribute children ever-born according to the same distribution Provides live births by period (year) and age of mother (single year)

Calculations: 3) ASFR

Principle: ASFR(a,t) = Births(a,t)/Women(a,t) Calculate ASFR, by age and period (single year) Merge age groups and periods as desired (age groups and periods add up in both numerator and denominator) Classic age group: 15-19, 20-24, …, 45-49 Periods: recommend 3 years: t-1 to t-3; t-4 to t-6 etc.

Examples: IPUMS Samples

Kenya 2009 census Available from IPUMS-international web site Data on children ever-born Data on maternal orphanhood Cross-tabulation of age of children by age of mother

Use MOMLOC and STEPMOM variables If STEPMOM>0, then MOMLOC = 0 (and thus this

child is not biological) If STEPMOM=0 and MOMLOC >0 then biological child

1) Age structure, Kenya 2009

0

10000

20000

30000

40000

50000

60000

10 15 20 25 30 35 40 45 50 55 60 65 70

Popu

latio

n

Age at census

Women

2) Fertility: children ever-born

0

1

2

3

4

5

6

7

8

10 15 20 25 30 35 40 45 50 55 60 65 70

Chi

ldre

n ev

er b

orn

Age of women

CEB

3) Maternal orphanhood

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45

0 5 10 15 20 25 30 35 40 45 50 55

Prop

ortio

n or

phan

Age of respondent

Mother died

4) Distribution of age of children by age of mother

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

0 5 10 15 20 25 30 35 40 45

Popu

latio

n

Age of children

12-19 20-29 30-39 40-49 50-59

Age of mother

Results

Kenya, 2009 census Level of fertility (TFR) Age pattern of fertility Fertility trends Comparison with 2008 DHS survey

Level of fertility, Kenya

5.09

6.04

7.52

4.56 4.89 4.71

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

0-4 years 5-9 years 10-14 years

TFR

Years before census

Census 2009 DHS surveys

Age pattern of fertility, Kenya (5 years before census)

0.000

0.050

0.100

0.150

0.200

0.250

10 15 20 25 30 35 40 45 50 55

ASF

R

Age of women

Census

Fertility trends, Kenya

0

1

2

3

4

5

6

7

8

9

1975 1980 1985 1990 1995 2000 2005 2010 2015

TFR

Period (year)

Census

Sources of bias

H1: Accurate age of mother H2: Accurate age of children H3: Accurate matching of children with mother H4: Same distribution of time since birth for children living with mother, children living elsewhere and children who died

Impact of women’s age misreporting (m= + 2, s= 5 years)

0.000

0.050

0.100

0.150

0.200

0.250

0.300

0.350

10 15 20 25 30 35 40 45 50 55

ASF

R

Age of women

Theoretical

H4: Mean age of children (duration since birth) Kenya, DHS, 2008

0

5

10

15

20

25

30

10 15 20 25 30 35 40 45 50

Age

of c

hild

ren

(yea

rs)

Age of mother (years)

At home Elsewhere Died

H4: Distribution of births by year before survey (Kenya, DHS,2008)

0%

1%

2%

3%

4%

5%

6%

7%

8%

9%

0 5 10 15 20 25 30 35

Perc

ent

Years before survey

Living at home Living elsewhere Deceased

Conclusion on Kenya case study

Level of fertility: acceptable for the past 5 to 6 years before census Age pattern of fertility: biased for older women Fertility trends:

Strongly biased for earlier periods More than with straightforward reverse survival

Application to other datasets: Kenya 1989

6.99 6.67 6.59

0

1

2

3

4

5

6

7

8

Kenya 1989

TFR

Census DHS UNPD

Kenya 1999

5.62

4.71 5.12

0

1

2

3

4

5

6

Kenya 1999

TFR

Census DHS UNPD

Zambia, 2000

5.64 5.90 6.15

0

1

2

3

4

5

6

7

Zambia, 2000

TFR

Census DHS UNPD

Burkina-Faso, 2006

7.80

6.00 6.26

0

1

2

3

4

5

6

7

8

9

Burkina, 2006

TFR

Census DHS UNPD

Cameroon, 2005

6.05

5.00 5.52

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

Cameroon, 2005

TFR

Census DHS UNPD

Mali, 2009

7.21 6.60 6.81

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

Mali, 2009

TFR

Census DHS UNPD

Fertility trends from series of successive censuses Use 3 censuses from Kenya

Full use of own-children method Restricted to past 5-6 years

Comparison of trends obtained by own-children with DHS, Kenya

0

2

4

6

8

10

12

1960 1970 1980 1990 2000 2010 2020

TFR

Period (year)

DHS

Census 1989 Census 1999

Fertility trends over past 6 years comparison with DHS, Kenya

2

3

4

5

6

7

8

9

1980 1985 1990 1995 2000 2005 2010

DHS

Census 1989 Census 1999

Limitations of Own-Children Method—both OWCH and OWCHMOM

Quality of data Age misreporting (children, mother)

Violations of hypotheses Independence between mortality, fertility, migration + magnitude of mortality & migration

Compared with reverse survival (GFR) More difficult to apply More subject to bias But provides an age pattern & series of TFRs

Need for triangulation

Check with levels and trends from other sources DHS surveys MICS surveys Other censuses

Conclusions

Reconstructing fertility levels and trends from census data is a challenge, but microdata are better than aggregated tables

Level of fertility in past 5 to 6 years Age pattern in past 5 to 6 years May lead to misleading fertility trends

Attempt worthwhile where census data are reliable Data quality: age reporting, mother location Potential biases: low mortality & migration

IPUMS MOMLOC is better than EASWESPOP Both OWCH & OWCHMOM are sensitive to:

data quality mortality inputs

But so is Reverse Survival

Lessons learned 1.  MOMLOC is great 2.  OWCHMOM is worth considering 3.  It’s the mortality, stupid!