United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14-17 November 2011 Evaluation of Fertility Data Collected from Population.

United Nations Sub-Regional Workshop on Census Data EvaluationPhnom Penh, Cambodia, 14-17 November 2011

Evaluation of Fertility Data Collected from Population Censuses

United Nations Statistics Division


For two items that can be collected to obtain fertility statistics in census:

• Children ever born

• Recent births

Discuss

• What information to collect

• What fertility indicators can be derived

• Possible quality issues related to each question

• Methods of data evaluation

Outline of the presentation


Children ever born – what information to collect?

How many children has [this woman] had in her lifetime? (a) “Total number of sons living in the household”;

(b) “Total number of sons living elsewhere”;

(c) “Total number of sons born alive who have died before the census date”;

(d) “Total number of daughters living in the household”;

(e) “Total number of daughters living elsewhere”;

(f) “Total number of daughters born alive who have died before the census date”.

incl. all live births

Could be elaborated into a number of questions

Asked to all women


When is it used?

• Widely used for over 50 years

• Important for countries without complete birth registration

• Also important for countries with complete birth registration

Study fertility by detailed socio-economic characteristics

Children ever born – the use of

What can we get?

Parity distributions Average number of children ever born

Age-specific fertility rate Total fertility rate (TFR)


Children ever born - Tabulation

Age of women

Number of children ever born

CEB not stated

Total women

Total children

born

0 1 … 14 15+

10-14

15-19

…

80-84

85+

NS

Total- Do not group the numbers of children, except for the last open category

- Distinguish children ever born not stated from no children


Children ever born – possible errors (1)

Group A: Errors because of misunderstanding the question:- Mortality error: reported only children living rather than ever born- Non-resident error: overlooking ever born children living elsewhere- Marriage error: women not including her children born from previous marriage/not

reporting children born out of wedlock

Group B: Errors because of respondents’ lapse of memory or neglect:- Memory error: forgot some children, especially from older women

Group C: Enumerators’ failure to reach individuals:- The not-at-home error: information provided by neighbors- Coverage error: omit an area or forgot to record the answer

Group D: Recording error:- Childless women mis-classified into parity not stated


• Not all errors can be detected

• However, many steps can be taken to find inconsistencies, to understand better the data quality and to provide information for improvement in the next census

Children ever born – possible errors (2)


Children ever born – quality assessment Methods (1)

- Initial assessment:

- Any missing values in children ever born data?

- Missing value for any relevant variables? (age of mother, sex of child, survival status of the child)

- Comparing children ever born and children surviving data



- For tabulated data:- Sex ratio of children ever born consistent with national average of sex

ratio at birth?- Check whether women with parity “not stated” are childless instead

(El-Badry method)- Plausibility of data – graphics help and best with other sources

- Average children ever born should increase with age (Group B error) – with constant or declining fertility assumption

- Proportion of women by parity, for older age groups- Time plot of mean number of children ever born, based on

multiple data sources (cohort analysis)- Age-specific fertility rates- TFR


P/F ratio method• Usually used to adjust current fertility data derived from recent births

question

• May also be used to assess the quality of both current (derived from recent births) and life time fertility data (derived from children ever born)

• Certain assumptions (discuss in more details later)



Children ever born – quality assessment examples (1) Any missing/implausible value for children ever born data?

Source: Estimation of fertility from the 2001 South Africa census data, Tom A Moultrie & Rob Dorrington, Centre for Actuarial Research, University of Cape Town


Children ever born – quality assessment examples (2)Comparing sex ratio at birth

Data source: graph produced based on data from United Nations Demographic Yearbook

Sex ratio of children ever born by age of mother, China 2000

1

1.02

1.04

1.06

1.08

1.1

1.12

1.14

1.16

1.18

15-19 20-24 25-29 30-34 35-39 40-44 45-49

Age of mother


Children ever born – quality assessment examples (3): Plot mean number of children ever born by age of women



Mean children ever born for Cambodia

0.00

1.00

2.00

3.00

4.00

5.00

6.00

Age

Avera

ge C

EB

1998 census

2008 census

2000 DHS

2005 DHS

1998 DHS

Children ever born – quality assessment examples (4): mean children ever born from multiple sources



Children ever born – quality assessment examples (5): parity distribution for age group 45-49

High level of 0 parity for older age group: 1950 and 1970 censuses: possible combined group of not stated parity with 0 parity group

Flat curve: probably some form of mis-reporting, seems to be improving over time

Mexican fertility survey: shape of the curve more plausible (although with small sample sizes)

Source: Child survivorship estimation: methods and data analysis, Griffith Feeney, Asian and Pacific Population Forum, Vol. 5, Nos. 2-3, 1991


• Year = census yr – (age – 25)

• 1960 and 1970 censuses: an increase of fertility

• Erroneous data from 1980 census (conclusion was reached after comparing with data from other surveys)

Children ever born – quality assessment examples (6): Cohort analysis of mean number of children ever born

Source: Child survivorship estimation: methods and data analysis, Griffith Feeney, Asian and Pacific Population Forum, Vol. 5, Nos. 2-3, 1991


Children ever born – quality assessment examples (7): Comparing age specific fertility rates

Data source: graph produced based on data from United Nations Demographic Yearbook and Measure DHS country report

Age specific fertility rates, Cambodia

0.0

50.0

100.0

150.0

200.0

250.0

15 - 19 20 - 24 25 - 29 30 - 34 35 - 39 40 - 44 45 - 49

Age

AS

FR

1998 DHS

2000 DHS

2005 DHS

2008 census

2010 DHS


Children ever born – quality assessment examples (8): Comparing total fertility rates

Data source: graph produced based on data from United Nations Demographic Yearbook and Measure DHS country report

Estimates of total fertility rates, Cambodia

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

1998 DHS 1998 census 2000 DHS 2005 DHS 2008 census 2010 DHS


Children ever born – quality assessment examples (9): El-Badry method

Age Female pop. Childless women (%)Women who have unknown number of children

(%)

15+ 1,483,895 20.1 9.7

15-19 250,329 63.7 32.2

20-24 229,655 35.5 13.0

25-29 214,467 11.4 4.9

30-34 187,348 4.8 2.6

35-39 135,551 3.5 2.2

40-44 97,537 3.5 2.2

45-49 75,526 3.3 2.3

50-54 76,100 3.8 2.4

55-59 50,817 4.0 2.4

60-64 53,775 4.6 3.0

65+ 110,062 4.7 4.1

Unknown 2,728 3.8 75.1

Data source: United Nations Demographic Yearbook, Burundi 1990


Children ever born – quality assessment examples (9): El-Badry method (cont.)

Data source: United Nations Demographic Yearbook, Burundi 1990


0

5

10

15

20

25

30

35

0 20 40 60 80

% women childless

% w

om

en

sta

tele

ss

ch

ildre

n

Y Predicted Y Linear (Predicted Y)

% women with not stated # of children = -0.49 + 0.48 * % childless women (linear regression)

%3248.1

48.0

32% of the real childless cases were reported as “number of children not given”

El-badry method: Failure of enumerators to make entries of zero: Errors in recording childless cases in population censuses, JASA, Vol. 56, No. 296, 1961

Children ever born – quality assessment examples (9): El-Badry method (cont.)

real childless women = observed childless women / (1-32%)


Children ever born – quality assessment examples (10)Using the P/F ratio method

Rational: - Compare cumulative fertility level derived from current fertility data F(trusting the

distribution but not level) and life time fertility data P (trusting the overall level but assumes under-reporting varies by age)

- The method is typically used to adjust current fertility level (may be generated from recent births question in census or birth data from civil registration)

- However the method is also used to assess the quality of children ever born data and sometimes, the age reporting of mother

Current fertility data: • Births in the last 12 months (from censuses or civil registration)

Life time fertility data:• Children ever born in the life time of women


Assumptions:

• Constant proportion of under-reporting of current fertility for all age groups

• Increasing under-reporting of parity (children ever born) by age of women

• Constant fertility (relaxed by a modification of the original P/F ratio method)

Children ever born – quality assessment examples (10)Using the P/F ratio method (cont.)


Children ever born – quality assessment examples (10): P/F ratio method (cont.)

Data source: Manual X, Bangladesh 1974 census

Typical P/F ratio, relative good data


Typical “look” of P/F ratios:

- Similar level of P/F ratios for age groups 20-24, 25-29 and 30-34

- P/F ratios becomes smaller for older ages

Deviation from the above typical pattern: indicates either violations of the assumptions or different patterns of under-reporting

- Example 1: a rising trend in the P/F ratios by age of women: fertility could have been decreasing in the past

- Example 2: a declining trend in the P/F ratios by age of women: fertility could have been increasing or that reported data on children ever born suffer from progressively increasing omissions of children as age of women increases

- Example 3: large fluctuation in P/F ratios may reflect either differential coverage by age or selective age misreporting of women

Children ever born – quality assessment examples (10): P/F ratio method (cont.)


Recent births - basics

How?

Date of birth of last child born alive (preferred question)

OR

Births in the last twelve months (to a woman) or in the household

However,

The questions are subject to under-reporting of births

Age of women during the census to be adjusted to age at giving birth (usually -0.5 year)

What can we get?

Age specific fertility rate and TFR


Recent births – Possible errors

Reporting errors:

Enumerator’s error

“reference period error”: uncertain of the date of birth vs the reference period

Proxy respondent

Births reported not including:

Women had a birth recently but died before the census

Household had a birth recently but the household dissolved before the census

Not significant in most cases, however could become an issue when many deaths occurred in a short period (HIV/AIDS)


Recent births – assessment (1)Methods

Initial assessment:any missing values in data? (month/date/year of births)

Missing data for any relevant variables? (age of mother, sex of child, survival status of the child)

For tabulated data:

Sex ratio at birth consistent with national average?

Plausibility of data – graphics help and best with other sources

Age-specific fertility rate

TFR


Recent births – assessment (1)Methods (cont.)

- Compare with civil registration data on live births

- P/F ratio method to compare with children ever born data for coverage


Recent births – assessment (2): any missing values?

Source: Estimation of fertility from the 2001 South Africa census data, Tom A Moultrie & Rob Dorrington, Centre for Actuarial Research, University of Cape Town


Recent births – assessment (3): Comparing age-specific fertility rate patterns

Source: Graph created using data from United Nations Demographic Yearbook; 2000 Census of Population, Social and Economic Characteristics of Population, State Institute of Statistics, Turkey; Measure DHS Final Report Turkey 1998 and 2003

Age specific fertility rate, Turkey

00.020.040.060.08

0.10.120.140.160.18

0.2

15 - 19 20 - 24 25 - 29 30 - 34 35 - 39 40 - 44 45 - 49

Age

2000 recent births

2000 PF ratio indirect method

2003 DHS direct

1998 DHS direct


Recent births – assessment (4): Comparing TFR

Source: Graph created using data from United Nations Demographic Yearbook; 2000 Census of Population, Social and Economic Characteristics of Population, State Institute of Statistics, Turkey; Measure DHS Final Report Turkey 1998 and 2003

TFR, Turkey

0

0.5

1

1.5

2

2.5

3

3.5

2000 recent births direct 2000 PF ratio indirect method 2003 DHS direct 1998 DHS direct

Source


Recent births – assessment (6): Comparing TFR

Yes under-estimated: but may reflect recent fertility pattern; adjusting for under-count and will be useful for sub-national and small area fertility estimates.

How to adjust: use household survey data

Only source for small area

Date of last birth data better quality than births in the last 12 months


Illustration of using MortPak FERTPF for indirect estimates of fertility – input data


Illustration of using MortPak FERTPF for indirect estimates of fertility – results


Thank you!

United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14-17 November 2011 Evaluation of Fertility Data Collected from Population.

Documents

total women total children

numbers of children

evaluation of fertility

c total number of sons

f total number of daughters

number of questions

ns total

memory error