492

i
ABSTRACT
Turkey has a considerably long history of nationwide demographic and health
surveys. Starting from 1963 Turkish Demographic Survey to 2003 Turkey
Demographic and Health Survey (TDHS-2003) eleven national surveys had been
carried out. Eight of these surveys were done by Hacettepe University Institute of
Population Studies (HUIPS). The last three of the surveys are based on the
Demographic and Health Survey (DHS) phase-3 program.
Despite the importance of its content and widely used characteristics of the TDHS
data, the data quality are not evaluated in a broad sense in terms of the quality of the
data having direct effect on fertility and mortality indicators. To evaluate the data
quality of TDHSs focusing on special variables effective on mortality and fertility
rates is aimed at this study. While discussing the overall quality of the data, the
impact of the quality on fertility and mortality rates are also aimed to be evaluated.
By using the simulations the effect of the data quality problems on Total Fertility
Rate, Infant Mortality Rate, Child Mortality Rate and Under-Five Mortality Rate are
evaluated. Assessing the data quality of DHS gives an idea about the common errors
faced. The results are discussed to give suggestions for future surveys.
The overall data quality at TDHS seems in good condition. Although age heaping
and digit preference problems are seen at the data, they are at tolerable levels. On the
other hand, the results of the simulations and estimations indicate that the problems
at data quality have no clear impact on the fertility and mortality rates. The results
indicate that regional and residential differences are seen in terms of the quality of
the data. The quality of data seems better at urban areas than rural. As the overall
data quality of the studied variables are high; among three surveys TDHS-1993 and
TDHS-2003 have higher data quality as compared to TDHS-1998.
ii
ÖZET
Türkiye ulusal düzeyde gerçekleştilien nüfus ve sağlık araştırmaları konusunda
oldukça uzun sayılabilecek bir geçmişe sahiptir. 1963 Türkiye Nüfus
Araştırmasından 2003 Türkiye Nüfus ve Sağlık Araştırmasına değin onbir ulusal
araştırma gerçekleştirilmiştir. Bu araştırmaların sekiz tanesi Hacettepe Üniversitesi
Nüfus Etütleri Enstitüsü (HÜNEE) tarafından yürütülmüştür. Bu araştırmaların son
üç tanesi Nüfus ve Sağlık Araştırmaları (DHS) programı temel alınarak
gerçekleştirilmiştir.
TNSA verisinin içerik önemi ve sıklıkla kullanımına rağmen doğurganlık ve
ölümlülük göstergeleri üzerinde doğrudan etkisi olan değişkenler geniş çaplı
incelenmemiştir. Bu çalışmada TNSA’nın veri kalitesinin ölümlülük ve doğurganlık
hızları üzerinde etkisi olan seçilmiş değişkenler üzerine odaklanılarak incelenmesi
amaçlanmıştır. Verinin genel kalitesinin tartışılmasının yanı sıra doğurganlık ve
ölümlülük hızları üzerine olan etkisi nin ölçülmesi de hedeflenmiştir. Simülasyonlar
kullanılarak, veri kalitesi sorunlarının Toplam Doğurganlık Hızı, Bebek Ölüm Hızı,
Çocuk Ölüm Hızı ve Beş Yaş Altı Ölüm Hızı üzerindeki etkisi değerlendirilmiştir.
TNSA’nın veri kalitesinin değerlendirilmesi karşılaşılan genel sorunlar hakkında bir
fikir verecektir. Sonuçlar gelecekte yapılacak olan araştırmalar açısından öneri
verecek şekilde incelenmiştir.
Genel olarak TNSA veri kalitesi iyi durumdadır.. Yaşa ilişkin değerlendirmelerde
yaş yığılması ve basamak tercihi gibi sorunlar görülmesine rağmen sorunların
boyutları kabul edilebilir düzeylerdedir. Simülasyon ve diğer hesaplamalar veri
kalitesi problemlerinin doğurganlık ve ölümlülük hızları üzerinde çok net bir
etkisinin olmadığını göstermektedir. Sonuçlar, bölgesel ve yerleşim yeri tipi
bağlamında farklılıkların olduğunu, genel olarak kentten elde edilen bilginin kıra
göre daha iyi olduğunu ortaya çıkarmaktadır. Her üç araştırma için çalışılan
değişkenler için genel veri kalite düzeyi yüksek olmakla birlikte bu araştırmalar
arasında TNSA-1993 ve TNSA-2003’nın veri kalitesi TNSA-1998’e göre daha
yüksektir.
iii
ACKNOWLEGDEMENTS

v
TABLE OF CONTENTS
ABSTRACT .......................................................................................................... i
ÖZET.................................................................................................................... ii
ACKNOWLEDGEMENTS ................................................................................ iii
TABLE OF CONTENTS ..................................................................................... v
LIST OF TABLES ............................................................................................ viii
LIST OF FIGURES .......................................................................................... xiii
I. INTRODUCTION AND OBJECTIVES ......................................................... 1
I.1. Introduction ............................................................................................... 1
I.2. Objectives .................................................................................................. 4
I.3. Contributions of the Study ......................................................................... 6
I.4. Organization of the Study........................................................................... 6
II. LITERATURE REVIEW ............................................................................. 10
III. DATA AND METHODOLOGY ................................................................. 28
III.1. THE HISTORY OF THE DEMOGRAPHIC SURVEYS
IN THE WORLD ............................................................................................ 28
III.2. THE HISTORY OF THE DEMOGRAPHIC AND
HEALTH SURVEYS...................................................................................... 31
III.3. THE HISTORY OF THE NATIONWIDE DEMOGRAPHIC
SURVEYS IN TURKEY ................................................................................ 34
III.3.1. Demographic Surveys carried out by HUIPS .................................. 35
III.3.2. Turkey Demographic and Health Surveys ....................................... 37
III.4. DATA SOURCES .................................................................................. 39
III.5. METHODOLOGY ................................................................................. 42
III.5.1. The Assessment of Data Used to Determine Eligibility for the
Individual Interview ................................................................................... 42
III.5.1.1. The Assessment of the Data at Household Interview ............... 43
III.5.1.1.1. Household Interview Results............................................ 43
vi
III.5.1.1.2. The Quality of Age Reporting in Household
Questionnaire ................................................................................... 44
III.5.1.1.3. Boundary Effects ............................................................. 48
III.5.1.1.4. The Household Residency ................................................ 50
III.5.1.2. The Assessment of Age Data in Individual Questionnaire ....... 51
III.5.1.2.1. Digit Preference .............................................................. 52
III.5.1.2.2. Imputation at the Age Data .............................................. 52
III.5.2.. The Assessment of the Quality of the Birth History Data ............... 52
III.5.2.1. The Quality of Birth Related Data ........................................... 53
III.5.2.1.1. Completeness of the information of
Birth Dates of the Children ................................................................ 53
III.5.2.1.2. The Displacement of Children’s Birth Dates .................... 54
III.5.2.1.3. Coverage of Live Births ................................................... 56
III.5.2.2. The Quality of Death Related Data .......................................... 56
III.5.2.2.1. Date of Birth Data ........................................................... 57
III.5.2.2.2. Age at Death Data ........................................................... 58
III.5.2.3. The Impact of Data Quality on Demographic Rates ................ 59
III.5.2.3.1. Fertility Impact of Data Quality ...................................... 59
III.5.2.3.2. Mortality Impact of Data Quality .................................... 60
IV. THE ASSESSMENT OF DATA USED TO DETERMINE
ELIGIBILITY FOR THE INDIVIDUAL INTERVIEW ................................. 63
IV.1. The Assessment of the Data at Household Interview.......................... 63
IV.1.1. Household Interview Results ...................................................... 63
IV.1.2. The Quality of Age Reporting in Household Questionnaire ........ 69
IV.1.3. Boundary Effects ........................................................................ 88
IV.1.4. The Household Residency........................................................... 91
IV.2. The Assessment of Age Data in Individual Questionnaire .................. 97
IV.2.1. Digit Preference ......................................................................... 97
IV.2.2. Imputation at the Age Data .......................................................100
vii
V. THE ASSESSMENT OF THE QUALITY OF THE BIRTH
HISTORY DATA ..............................................................................................109
V.1. The Quality of Birth Related Data. .....................................................109
V.1.1. Completeness of the information of Birth Dates of
the Children .........................................................................................109
V.1.2. The Displacement of Children’s Birth Dates ...............................123
V.1.3. Age Heaping ..............................................................................132
V.1.4. Miscalculation of Year of Birth ...................................................138
V.1.5. Coverage of Live Births ..............................................................141
V.2. The Quality of Death Related Data. ...................................................145
V.2.1. Date of Birth Data ......................................................................145
V.2.2. Age at Death Data ......................................................................150
V.2.3. Accuracy of the Data ..................................................................159
V.3. The Impact of Data Quality on Demographic Rates. ..........................164
V.3.1. Fertility Impact of Data Quality .................................................164
V.3.2. Mortality Impact of Data Quality................................................169
V.3.2.1.The Impact of Boundary Effect and Sleeping Away
Exclusion on U5MR .........................................................................169
V . 3 .2.2. The Impact of Heaping of Deaths at Twelve Months of Age,
on IMR and CMR Estimates. ............................................................174
VI. CONCLUSIONAND DISCUSSION ..........................................................181
VII. REFERENCES ..........................................................................................197
VIII. ANNEXES ................................................................................................208
viii
LIST OF TABLES
Table III.1. Sample Size and Completed Interviews at Demographic and Health
Surveys, Turkey 1993, 1998, 2003 ..................................................... 39
Table III.5.1.1.2.1. Whipple Index Score for Estimating Reliability of Age Data . 45
Table IV.1.1. Household Response Rates and Percent Distribution of
Household Result Codes by Region and Type of Place of Residence,
TDHS-1993................................................................................... 66
Table IV.1.2. Household Response Rates and Percent Distribution of
Household Result Codes by Region and Type of Place of Residence,
TDHS-1998................................................................................... 67
Table IV.1.3. Household Response Rates and Percent Distribution of
Household Result Codes by Region and Type of Place of Residence,
TDHS-2003................................................................................... 68
Table IV.1.2.1. De Facto Age Distribution of TDHS-1993 ................................... 71
Table IV.1.2.2. De Facto Age Distribution of TDHS-1998 ................................... 72
Table IV.1.2.3. De Facto Age Distribution of TDHS-2003 ................................... 73
Table IV.1.2.4. Household Age and Sex Ratios and Myers, Bachi, Whipple and
United Nations Indices for Household Data by Region and Type of
Place of Residence, TDHS 1993, 1998 and 2003 ...................... 84
Table IV.1.2.5. Myers, Bachi, Whipple Indices for Household Data by Demographic
Characteristics of Respondent whom the Household Interview is
Completed, TDHS 1993 .............................................................. 86
Table IV.1.3.1. Indices of Age Eligibility Distortion Based on Household Data by
Region and Type of Place of Residence, TDHS 1993, 1998 and
2003 ......................................................................................... 89
ix
Table IV.1.3.2. Indices of Age Eligibility Distortion Based on Household Data by
Demographic Characteristics of Respondent whom the Household
Interview is Completed, TDHS 1993, 1998 and 2003 .................. 91
Table IV.1.4.1. Evaluation of the Sleeping Away Exclusion of the Female Usual
Residents, TDHS-1993. .............................................................. 94
Table IV.1.4.2. Evaluation of the Sleeping Away Exclusion of the Female Usual
Residents, TDHS-1998. .............................................................. 95
Table IV.1.4.3. Evaluation of the Sleeping Away Exclusion of the Female Usual
Residents, TDHS-2003. .............................................................. 96
Table IV.2.1.1. Percent Distribution of Women 20-49 by Reported Terminal Digit of
Age (Individual Questionnaire) and Myers index by Region, Type of
Place of Residence and Education of Woman, TDHS 1993, 1998 and
2003. ........................................................................................... 99
Table IV.2.2.1. Percent Distribution of the Completeness of the Date of Birth and
Age Information by Region and Type of Place of Residence, TDHS
1993, 1998 and 2003. .................................................................103
Table IV.2.2.2. Percent Distribution of the Completeness of the Date of Birth and
Age Information by Demographic Characteristics of Women, TDHS
1993, 1998 and 2003 ..................................................................105
Table V.1.1.1. Percent Distribution of Children Born by Completeness of
Information on Date of Birth by Region and Type of Place of
Residence, TDHS 1993, 1998 and 2003 .....................................113
Table V.1.1.2. Percent Distribution of Children Born by Completeness of
Information on Date of Birth by Demographic Characteristics of
Women, TDHS 1993, 1998 and 2003 .........................................114
Table V.1.1.3. Percentage of Children with Complete Information on Year and
Month of Birth by Number of Years since Birth of Child by Region
and Type of Place of Residence, TDHS 1993, 1998 and 2003 ....117
x
Table V.1.1.4. Percentage of Children with Complete Information on Year and
Month of Birth by Number of Years since Birth of Child by
Demographic Characteristics of Women, TDHS 1993, 1998 and
2003...........................................................................................118
Table V.1.1.5. Percentage of Children with Complete Information on Year and
Month of Birth by Number of Years since Birth of Child by Survival
Status and Sex of Child and Time Period of Interviewer at the Field,
TDHS 1993, 1998 and 2003 .......................................................121
Table V.1.2.1. Number of Questions in Ever-Married Woman Questionnaires
that Depend on Children’s Year of Birth, TDHS 1993, 1998 and
2003...........................................................................................124
Table V.1.2.2. Median Minutes to Complete the Ever-Married Woman Questionnaire
by the Number of Children Born in Last Five Years prior to Survey,
TDHS-2003. ..............................................................................124
Table V.1.2.3. Number of Births by Calendar Years and Birth Year Ratios for 4, 5
and 6 Years prior to the Survey by Region and Type of Place of
Residence, TDHS 1993, 1998 and 2003 .....................................127
Table V.1.2.4. Number of Births by Calendar Years and Birth Year Ratios for 4, 5
and 6 Years prior to the Survey by Demographic Characteristics of
Women, TDHS 1993, 1998 and 2003 .........................................128
Table V.1.2.5. Number of Births by Calendar Years and Birth Year Ratios for 4, 5
and 6 Years prior to the Survey by Survival Status and Sex of Child
and Time Period of Interviewer in the field, TDHS 1993, 1998 and
2003...........................................................................................130
Table V.1.3.1. Age Ratios for Living Children by Single Year of Age by Region and
Type of Place of Residence, TDHS 1993, 1998 and 2003 .............134
Table V.1.3.2. Age Ratios for Living Children by Single Year of
Age by Demographic Characteristics of Women, TDHS 1993, 1998
and 2003 ......................................................................................136
xi
Table V.1.3.3. Age Ratios for Living Children by Single Year of Age by Sex of the
Child and Time Period of Interviewer in the Field, TDHS 1993, 1998
and 2003. .....................................................................................137
Table V.1.4.1. Percentage of Children Whose Month of Birth Falls in the Month of
Interview or Earlier by Region and Type of Place of Residence, TDHS
1993, 1998 and 2003 ....................................................................140
Table V.1.5.1. Average Number of Children Ever Born by Age of Mother by Region
and Type of Place of Residence, TDHS 1993, 1998 and 2003 ......143
Table V.1.5.2. Median Age at First Birth by Age of Woman at the Time of Survey
by Region and Type of Place of Residence, TDHS 1993, 1998 and
2003 .............................................................................................144
Table V.2.1.1. Percentage of Births with Incomplete Information on Date of Birth by
Survival Status by Region and Type of Place of Residence, TDHS
1993, 1998 and 2003. ...................................................................148
Table V.2.1.2. Percentage of Births with Incomplete Information on Date of Birth by
Survival Status by Demographic Characteristics of Woman, TDHS
1993, 1998 and 2003. ...................................................................149
Table V.2.2.1. Percentage of Deaths with Incomplete Information on Age at Death
by Calendar Period in Which the Birth Occurred by Region and Type
of Place of Residence, TDHS 1993, 1998 and 2003. .....................152
Table V.2.2.2. Percentage of Deaths with Incomplete Information on Age at Death
by Calendar Period in Which the Birth Occurred by Demographic
Characteristics of Woman, TDHS 1993, 1998 and 2003. ..............153
Table V.2.2.3. Total Reported Deaths and the Number of Deaths with Incomplete
Age at Death Information by Type of Defect in Information by Region
and Type of Place of Residence, TDHS 1993, 1998 and 2003 ......156
Table V.2.2.4. Total Reported Deaths and the Number of Deaths with Incomplete
Age at Death Information by Type of Defect in Information
by Demographic Characteristics of Woman, TDHS 1993, 1998 and
2003 .............................................................................................158
xii
Table V.2.3.1. Index of Heaping of Deaths at Twelve Months of Age by Region and
Type of Place of Residence, TDHS 1993, 1998 and 2003 .............161
Table V.2.3.2. Index of Heaping of Deaths at Twelve Months of Age
by Demographic Characteristics of Woman, TDHS 1993, 1998 and
2003. ............................................................................................162
Table V.3.1.1. Results of Simulations to Estimate the Effect of Lower Boundary,
Upper Boundary and “Sleeping Away” Exclusions on Total Fertility
Rate, TDHS 1993 .........................................................................166
Table V.3.1.2. Results of Simulations to Estimate the Effect of Lower Boundary,
Upper Boundary and “Sleeping Away” Exclusions on Total Fertility
Rate, TDHS 1998 .........................................................................167
Table V.3.1.3. Results of Simulations to Estimate The Effect of Lower Boundary,
Upper Boundary and “Sleeping Away” Exclusions on Total Fertility
Rate, TDHS 2003 .........................................................................168
Table V.3.2.1.1. Results of Simulations to Estimate the Effect of Lower Boundary,
Upper Boundary and “Sleeping Away” Exclusions on Under-Five
Mortality Rate, TDHS 1993 .......................................................171
Table V.3.2.1.2. Results of Simulations to Estimate the Effect of Lower Boundary,
Upper Boundary and “Sleeping Away” Exclusions on Under-Five
Mortality Rate, TDHS 1998 .......................................................172
Table V.3.2.1.3. Results of Simulations to Estimate the Effect of Lower Boundary,
Upper Boundary and “Sleeping Away” Exclusions on Under-Five
Mortality Rate, TDHS 2003 .......................................................173
Table V.3.2.2.1. Estimates of Infant and Child Mortality for the Five year Period
Preceding the Survey, Adjusted for Heaping of Deaths at Twelve
Months of Age, by Region and Type of Place of Residence, TDHS
1993, 1998, 2003. ....................................................................176
xiii
LIST OF FIGURES
Figure IV.1.2.1. Age Distribution of De Facto Household Population,
TDHS-1993 ............................................................................... 74
Figure IV.1.2.2. Age Distribution of De Facto Household Population,
TDHS-1998 ............................................................................... 74
Figure IV.1.2.3. Age Distribution of De Facto Household Population,
TDHS-20033 ............................................................................. 75
Figure IV.1.2.4. Myers Preference by Digit, TDHS-1993 ..................................... 77
Figure IV.1.2.5. Myers Preference by Digit, TDHS-1998 ..................................... 77
Figure IV.1.2.6. Myers Preference by Digit, TDHS-2003 ..................................... 78
Figure IV.1.2.7. Percentage of Women 10 to 60 Years,
TDHS 1993, 1998 and 2003...................................................... 87
Figure V.1.2.1. Percent Distribution of Births for Ten Years prior to Survey,
TDHS 1993, 1998 and 2003 .......................................................132
TO MY FAMILY:
WHICH I WAS BORN INTO,
WHICH I HAVE CREATED...
1
I. INTRODUCTION AND OBJECTIVES
I.1. Introduction
Fertility, mortality and migration are the three main study areas of demography.
Since the first studies considered as “demographic” made by academicians, scientists
or researchers; these subjects have been “sine qua non” (without which -there isnothing")
for most of the studies. As they are the main components of the population
change, most of the studies on demography are related with them. The need for
accurate, complete and up to date data is crucial for demographic studies. The vital
registration systems aim to record these basic demographic events regularly for the
whole population. In addition, both various small scale and the worldwide
demographic surveys are designed to get information on fertility, mortality and
migration. With statistical approaches, specific rates, proportions, ratios and various
indexes are developed on these subjects to evaluate the differences within and
between populations. In most of the developing countries, sample surveys and
censuses are the main sources used to estimate available demographic indicators
since the vital registration systems have both coverage and quality and reliability
problems.
As the surveys are the only source for various demographic events and rates; the
quality of the survey data is crucial. Not only the researchers who conduct these
surveys, but also the academicians and policy makers who evaluate the survey results
concern about the quality of the data. Reliable data gathered from field surveys is
valuable and important to estimate reliable indicators which are commonly used by
policy makers in the countries where registration systems have no ability to produce
these indicators. In addition, up to date and better quality data are constantly pursued
at the country level by governments, nongovernmental organizations and
international organizations to develop policies, programs and interventions, and to
2
monitor their accomplishments in improving the living conditions of different
populations. Moreover, at the international level, good quality data is needed to
monitor the goals and targets agreed at international forums and in particular for the
Millennium Development Goals (Loaiza, 2004).
In Turkey, besides censuses, nationwide surveys starting from 1963 Turkish
Demographic Survey to 2003 Turkey Demographic and Health Survey had been
carried out by School of Public Health, Ministry of Health and Hacettepe University
Institute of Population Studies (HUIPS). Surveys were designed to collect data on
various demographic and health subjects are the only sources for many demographic
indicators commonly used by not only social scientists but policy makers as well.
The information collected by the surveys is used to determine the demographic
situations at the time of survey and evaluate the trends and changes of the
demographic indicators within the survey periods. The main concern of these surveys
was to put on the demographic situation within the intercensal periods for Turkey.
These surveys were applied at the years ending 3 and 8 after 3 years of each census.
Nearly all of these surveys were a part of worldwide demographic and health surveys
and especially technical assistance was provided by the international agencies
(HUIPS, 2008, Macro International Inc, 2008). Therefore, at the critical stages of the
survey like sampling, questionnaire design, data entry, tabulation and dissemination
the experience of the international agencies helped the quality of the overall survey.
Surveys -because of their nature- carry certain problems with themselves depending
on their sample design, data collection, methodology, questionnaire design, wording
and sequence of the questions, interviewers’ training period, content and the topic
interested in. In addition, the errors may be large or small depending on the obstacles
to accurate recording which are present in the area concerned, the methods used in
compiling the data, and the relative efficiency which the methods are applied (UN,
1955). It is possible to minimize the effect of these errors on the quality of the results
of a survey; getting rid of them is impossible. Errors can be classified into two broad
categories; sampling and non-sampling errors.
3
In most of the developing countries there exists a gap for reliable data sets for the
precise estimation of demographic indicators. In countries where reliable
demographic data is scarce, the estimates based on the birth history module in
demographic and health surveys are particularly desirable to obtain important
demographic measures. (Goldman, et al. 1979). Likewise, demographic and health
surveys conducted in Turkey are accepted as the sources for accurate and reliable
results about population components. However, potential errors arise from the survey
data are the main misleaders for the data users as it is in other sample surveys. They
cannot be free from both sampling and non-sampling errors that may produce biased
results. In both Turkish and international literature, there are various studies about
the errors originated from respondent, interviewer and data collection tools in
censuses and surveys which are being studied with indexes like Myers and Whipple
(Al Abdel 1987, Albayrak 1991, Tungul 1995) However, these studies are limited in
a descriptive position in which there is only the amount of the deviations from the
expected age structure have been studied.
The last three of the surveys carried out by HUIPS is based on the Demographic and
Health Survey (DHS) phase-3. With Turkey Demographic and Health Surveys
(TDHSs) 1993, 1998 and 2003 -as a standard schedule- HUIPS published
preliminary reports followed by main report and summary reports in which various
topics covered by questionnaires are included. However, data quality of the study is
discussed in a limited way, not in line with its importance. Only basic data quality
tables were supplied in TDHS-93 and TDHS-98 main reports in Appendix D sections
(HUIPS 1994, 1999). Similarly, in TDHS-2003 main report, in addition to the tables
a brief explanation of the tables had found place for the first time (HUIPS, 2004).
However, these “basic” data quality tables are inadequate for DHSs as compared to
their value. On the other hand the data quality of TDHS-1993 was studied at the
further analysis report in 1997 by HUIPS staff (HUIPS, 1997). Hancıoğlu (1997) at
this report discuss the data quality of TDHS-1993 both on household and ever
married women levels.
4
Assessment of the data quality of the birth history and data used for direct estimation
of infant and child mortality is very important. DHS program carried out only in
developing countries and most of the developing countries have known to be famous
with having high fertility and infant and child mortality rates. Therefore the quality
of the data of these issues are more essential than any other sections in terms of
having reliable estimations which is valuable for policy makers to monitor the future
programs. In Turkey, the plans, programs, policies and estimations for resources
(human, financial) developed by the Ministry of Health and State Planning
Organization are based on these rates. In Official Statistical Program organized by
Turkish Statistical Institute (TURKSTAT) for the years 2007-2011, selected
indicators of TDHSs are accepted as “official statistics” under “health statistics” and
“demographic statistics” sections.
I. 2. Objectives
This study has three different but highly interrelated aims. The first one is to evaluate
the data quality of TDHSs. As the nationwide surveys carried out by HUIPS is wellknown
with their accurate and reliable results on various demographic and health
issues. However, it seems important to assess the data quality of TDHSs in order to
put forward the strong and weak points in terms of data quality. Same statistical tests
of data quality are going to be applied to all three TDHSs, so that the general quality
of the data sets can be assessed and a comparison between these surveys can be done.
Infant Mortality Rate (IMR) and Child Mortlity Rate (CMR) are seen as the critical
measures of the wellbeing of the children and a good proxy indicator of the overall
level of development of the country. Moreover, fertility rates are very important in
terms of the effects both on the quality and the quantity of the population. The
quality of the data used for the direct estimation of infant and child mortality (IMR
and CMR) and fertility is the primary concern of this study. However not only the
women but the data collected at household level is going to be studied on. As the
eligibility of ever married women for applying the women questionnaire is decided
5
on the information collected at the household list, the quality of age and usual
residency information is crucial for the data quality.
The focus of the data quality evaluation at ever married questionnaire is first on the
age and birth date data of the woman at the first section of the questionnaire and the
birth history section. Birth history data is the source for estimating the fertility and
infant and child mortality rates. This study is interested in the possible and observed
errors in birth history data in TDHSs. As the birth and death date of the child has
direct effect on fertility and mortality rates. The quality of the birth and death date is
aimed to be studied at this study.
To evaluate the effects of the errors on fertility and mortality rates is the second aim
of this study. Both the information at the household questionnaire and ever married
individual data is used to determine the effect of data quality on the mortality and
fertility rates. The impact of the displacement of the eligible women out of ages 45-
49 and the sleeping away exclusion of the women on Total Fertility Rate and Under
Five Mortality Rates are going to be studied. In addition, the effect of heaping on
12th month on age at death is going to be studied on the Infant Mortality Rate and
Child Mortality Rate.
The third aim of this study is to identify the errors that are inherit in TDHSs which
may indicate problems that need to be addressed or changes that need to be made in
future surveys. DHSs are alike with other many small/large scale field surveys in
terms of non sampling errors. The underlying reasons for the possible non-sampling
errors effecting data quality and solutions to free from these errors will be studied.
The possible reasons of the errors and the ways to overcome these errors are going to
be discussed at this study.
6
I.3. Contributions of the Study
The TDHSs 1993, 1998 and 2003 results and datasets are used not only by HUIPS
staff, but social scientists form different departments, universities and policy makers
as well. Despite its broad usage, the data quality of the surveys is not studied in
detail. This study will help the users to judge the survey results in a more correct
situation and the gap on assessing the data quality is going to be filled. The overall
quality of the data used for the fertility and mortality of children is aimed to be
studied.
Assessing the data quality of TDHS will both put forward the strong and weak sides
of the TDHSs in terms of data quality and give suggestions for future demographic
studies in the lights of results. The data quality problems according to the
questionnaire and fieldwork will be evaluated at this study. Therefore, the results of
this study will be helpful for the researchers who are going to carry future
demographic surveys.
The lessons learned at these surveys in terms of improving the data quality of the
further studies are also going to be discussed. The common data quality problems at
TDHSs and the proposals for the solution are aimed to be discussed.
I. 4. Organization of the Study
Study starts with the introduction chapter where the justification of the topic selected
for this thesis is mentioned at the first stage. The reasons to estimate the data quality
of TDHSs are discussed. In addition, the objectives and the contribution of the thesis
to the literature are discussed at this chapter too. Chapter ends with the organization
of the thesis which will also help the readers to understand the coverage of the thesis.
In Chapter II, the literature review on the data assessment of the surveys is discussed.
The former studies will be evaluated and the content and the contribution these
7
studies are presented. Chapter III is reserved for the data and the methodology in
which the history of the Demographic and Health surveys in the World and Turkey is
presented. In addition the structure of DHSs –the questionnaires used, the months
they carried on, the sampling design, etc. - specifically TDHSs is discussed here.
In Chapter IV, the assessment of the data used to determine the eligibility for the
individual interview is presented. The household interview results’ are evaluated in
terms of result codes. Age information of the members of the household list is going
to be studied in the household questionnaire. The main focus is the digit preference
which is evaluated with Myers Blended Index and Whipple Index. While examining
the quality of the age data at household survey, some important characteristics (age,
sex, relationship to the household, education level, etc) of the respondent with whom
the questionnaire is filled are going to be discussed.
In addition, the reported age of the women which has direct effect of the eligibility of
the women is discussed. The scope of error is going to be evaluated with upper
boundary and lower boundary effect indices. Moreover, by assessing the results for
the question on the information whether the women stayed last night at the house
which shows the extent of the problem of the exclusion of women who did not sleep
the night before the interview is discussed.
The assessment of age data in individual questionnaire is also aimed to be studied at
the fourth chapter. The methods on the age information at the individual data are
discussed at this part of the chapter. Myers and Whipple Indexes are estimated to
understand the extent of the digit preference. In addition to these calculations, five
year age group distortions are computed to understand the total picture of the age
distribution problem. In addition the extent of the imputation on age information is
discussed.
The quality of the birth history data is placed at Chapter V. The answers to the
questions: What is the extent of missing information of the birth dates of the
children? In what percentage is the imputation done? are aimed to be answered. The
8
problem of carrying the birth dates of the children out of the five year period which
gives the interviewer to escape from the workload of asking additional questions in
the next sections of the questionnaire about the children under five. This problem is
discussed in terms of the length of the section about the children under five. In
addition the duration of the interviewer in the field, and the comparison of
interviewers with other interviewers in terms of the quality of the data are going to be
done.
The digit preference and the age heaping problems in the birth history data are also
going to be discussed at Chapter V. The problem of miscalculation of year of birth
by either the interviewer, or the respondent is discussed. If the mother doesn’t know
the birth month of the children, either the respondent or the interviewer may
calculate the age of the children by easily subtracting the year of birth from the year
of interview. If the month of birth is not known and the age of the respondent is not
equal to the result of the subtraction of year of interview and year of birth then the
age is imputed during the data entry. The level that the imputation done is important
in terms of data quality of the TDHSs and assessed at this chapter.
On the other hand, the quality of the data used for the direct estimation of infant and
child mortality is also evaluated at Chapter V. The completeness of the date of birth
data and its accuracy is re-discussed (heaping, omission, etc.) at this chapter. The
completeness of the age at death data and the extent of the missing information are
going to be evaluated. Moreover at this chapter the accuracy of the data; the age
heaping problem especially on month 12 and its impacts on mortality estimates is
also going to be discussed.
The Impact of the Quality of Data on Fertility and Mortality Rates is discussed at the
last section of Chapter V. The impacts of the problems in data on rates are discussed
with giving references to each of the estimation. The last part of the chapter assesses
the simulations made which will show the effect of the possible data quality
problems on the Total Fertility, Under Five Mortality and Infant Mortality Rates.
9
Chapter VI. is reserved for discussion and conclusion. The results are evaluated to
discuss the solutions to reduce or eliminate these problems in the future. The overall
data quality estimations are assessed at this chapter and comments and results for
future studies are discussed. In addition, the efforts made on the field study to
decrease the errors on the data and a new approach to data entry which may have
effects on data quality is also discussed at this chapter.
10
II. LITERATURE REVIEW
The literature on the data quality of surveys has a long history at social sciences. The
researchers have always been interested with the quality of the data gathered by the
censuses, registration systems and surveys. The reliability of the results has always
been questioned and the level of sampling and non-sampling errors are tried to be
estimated. Several statistical methods were structured to estimate the level of
sampling errors for different type of surveys, censuses and registration system.
Depending on the type of sampling, special formulas and techniques were used to
assess the sampling errors. In addition, the non sampling errors, caused by the
interviewer, the respondent or the questions themselves are also widely studied and
basic techniques are developed to put forward the level of errors.
Various studies on the quality of the data collected by the local and international
studies developed and applied by the researchers especially in USA are seen at the
literature. After the Second World War the importance of the knowledge on the
population gained importance in terms of developing policies. Various studies were
started to be carried out on special issues on population interested. Both the surveys
directly conducted by Census Bureau and other American companies especially on
the topics of fertility and mortality in developing world is well known about their
technical papers on the quality of the data gathered from the field. Starting from the
quality of survey design in pure statistics concerns to quality of variables used for the
estimation of rates and ratios are studied thoroughly.
It is expected and seen that UN has always had great interest on the information on
population especially of the developing and undeveloped countries. Therefore, the
data gathered by censuses, registration systems and surveys have an importance in
policies and programs offered and followed by UN. Data quality of the data is
focused on any kind of statistics had been specifically studied with one of the series
11
of manuals for the estimating the population. In 1955, one of the second series of
manuals; Manual II “The Methods of Appraisal of Quality of Basic Data for
Population Estimates” was published by Population Branch of the United Nations
Bureau of Social Affairs (UN,1967). Manual concerns the data quality of the census
enumerations of the total population and sex and age groups. The basic aim of the
manual is to assist governments in improving the quality of official population
estimates. As the methods for estimating the data quality of demographic data are
gathered together for the use of social scientists; this study can be considered as one
of the cornerstone studies on the data quality.
UN followed the processes in the methodology of data quality and published reports
on the quality of data of World Fertility Surveys conducted especially on the
undeveloped and developing areas of the world.
During the WFS program, various research papers and methodological papers/studies
were published. Most of the studies were the reflection of the concerned points of the
program. The quality of the data gathered from the WFSs was also another important
interest area of the program. Rindfuss; Bumpass and Palmore published an article in
“Demography” in 1987. They discussed the ways in which the restricted fertility
histories produce a biased sample of births. In addition, they evaluated “the effect of
the restriction by using a high-quality data set that does not contain the usual
restriction” (Rindfuss et al. 1987). They used the 1974 Korean WFS dataset and
discussed the information on contraception and breastfeeding for every birth interval.
Thus, they could be able to analyze the determinants of birth intervals with and
without the WFS restrictions to examine the degree of bias they introduce. They tried
to find answers to the level of the errors of the selection of last closed and open
intervals which lead to biased estimates of the levels of contraceptive use and
breastfeeding duration. In addition, they focused on the bias sourced by this selection
by which the findings regarding the structure of relationships between these
parameters and other variables of concern.
12
When the birth histories started to be used in surveys, some criticisms was made on
the quality of the data collected with birth histories especially at the undeveloped and
developing countries. Potter (1997) made one of the remarkable critiques on birth
history data of El Salvador (WFS-1973), and Bangladesh (WFS-1961/62 and
National Impact Survey-1969). By using simulations with the data, he tried to put on
the actual fertility in these countries at the time of survey and the estimated ones
from the survey. He noticed that the estimations based on birth history data are too
sensitive to the age misreporting and will lead to an artificial decline in fertility rates.
The scientific reports published at the WFS project covered the assessment of the
data quality in participating countries. In addition to these single country reports
evaluating data quality, a comparative report was published by using the data from
41 countries to assess the quality of the WFS data (Goldman , et al, 1985). At this
study, a few general types of error are recognized; omission of events -like live births
or unions- and household members; and misreporting of dates of events. They
discuss the level of the sampling errors, the errors originated from the wording of the
questions and the effects of imputation at the results. They conclude that the
imputation of the data at WFSs don’t have a clear effect on the general results.
The data sets of different WFS countries were studied by researchers on special
interest areas. Chidambaram and Pullum focused on the sensitivity of estimated
fertility levels and trends to alternative interpretations of the responses by using the
data from the 1975 Bangladesh Fertility Survey (BFS). Some simple mathematical
models were used to evaluate the consequences of an incorrect interpretation and
they also offered guidelines for a correct interpretation. They focused on the 'years
ago' responses used in WFS questionnaires. They indicate that there will be a
problem that will occur” indirectly when the respondent and/or interviewer
essentially mimic the imputation procedure by first estimating an age or elapsed time
and, then converting this to a date.” (Chidambaram and Pullum, 1981)
Before the DHS program is developed and applied, many evaluations of the previous
surveys were done to increase the quality of the DHS and further surveys in the
13
future. Cleland (1986) published an article in which he reviewed the experience of
the Contraceptive Prevalence Survey (CPS) and the World Fertility Survey (WFS),
and attempted to identify their major implications for similar future projects,
particularly the Demographic and Health Survey (DHS). The lessons learned from
DHS’s predecessors are thoroughly absorbed and with this article personal view of
these lessons are presented by Cleland and some key choices that the DHS must face
were told about.
An example evaluating the WFS experience is taken into consider by Pullum,
Harpham and Ozsever (1986). They published an article titled: “The Machine
Editing of Large-Sample Surveys: The Experience of the World Fertility Survey”
focused on the various costs and benefits attached to the machine editing of data
which accompanied the preparation of standard computer files for each country.
Although their paper was based exclusively on WFS experience, and functions in
part as a summary and appraisal of that experience, they also intended to have some
value for the planning of similar surveys in the future. They deal with six countries’
data sets selected on the point that they have the early unedited raw data file and
constructed a ‘dirty’ Standard Recode File and compare it with the ‘clean’ Standard
Recode to understand the effect of machine editing. They used 25 indicator variables
used in tabulation in country reports. The results indicated that percentage of the
cases lost in the matching is very low except Ghana (6%) because most of the
structural changes were already done in earlier stages of the survey. Out of 147
different occurrences of these 25 indicators in these 6 countries only 12 of them
seems to have different distribution more than 1%. Changes appeared to be least
likely for numbers of children ever born and living’ fertility desires and background
variables. In addition they found that the TFR estimated both from dirty and clean
data are quite close except Ghana.
Another study on the experience of WFS is on the quality of the birth history data
collected by WFS. At this study written by Hobcraft, Goldman and Chidambaram in
1982, the results of the rates estimated directly by birth history and with the aid of
the conventional P/F procedure is compared. They stressed that “when complete
14
birth histories are available, the basic fertility rates themselves should always be
examined and the P/F procedure should be viewed as one of a series of measures
which aid in the interpretation of fertility data” (Hobcraft, et al. 1982). They also
indicated that it is more natural and much simpler to analyze the birth histories by
using period rates for cohorts rather than more conventional period rates for age or
duration groups.
Pullum (1991) continued his studies on data quality of demographic surveys with the
assessments with DHS datasets. He attempted to illustrate general difficulties
associated with the reporting of ages and dates in developing countries by focusing
on Pakistan DHS. He indicated that misreporting of the age is a common problem in
Pakistan. He advises that, if the reports of the surveys are published without
concerning the quality of the age data, the results will be misleading for the users. He
focused on the birth history section of the data and developed a model that shows the
extent and severity of the transfer of infants and one-year olds into the later ages of
childhood. Under this model, he mentioned that the reported number of infants is
plausibly adjusted to correspond with the reported number of births (Pullum, 1991).
Another study done by Pullum is the Methodological Reports 5 published by Macro
International Inc. which deals with the quality of age and date reporting in DHS
surveys (Pullum, 2006). Most of the indicators produced by DHS surveys depend on
accurate reporting of ages of women and children, as well as dates when events
occurred are assessed with this report. The 141 DHS surveys conducted from 1985 to
2003 in 66 countries are examined. The center of the attention of the report is
measuring the levels and patterns of four kinds of potential errors: incompleteness,
heaping or digit preference, transfers across boundaries, and inconsistencies between
successive surveys. Household and individual data from nearly all of the DHS
surveys are used at this report to identify evidence of misreporting of ages and dates.
Report focuses on the levels of incompleteness, digit preference, and transfers across
specific boundaries such as ages 5, 15, and 50. It measures the heaping of age at
death at 12 months with a logit regression calculation with the births in the ten years
before the survey.
15
DHS program made evaluations between the stages of the program. In 1996
Marckwardt and Rutstein prepared a working paper “Accuracy of DHS-II
Demographic Data: Gains and Lose in Comparison with Earlier Surveys”
(Marckwardt and Rutstein, 1996). The Datasets from DHS-II program are compared
with the data collected in the earlier WFS and DHS-I programs. They aimed to
measure any improvements in the probable accuracy of demographic measures, as
well as detect any problem areas where quality may have deteriorated, or not
changed. They examined the accuracy of two basic demographic measures, current
fertility rates and current infant mortality rates, as measured in the second round of
DHS surveys. They deal with issues like Distortions at the boundaries of age
eligibility and Distortions in Reporting De facto status in household list. In addition,
they put forward the improvements at the data quality of birth history. They discuss
the possible effects of the distortions, displacements at the birth history data to
fertility and mortality rates calculated directly with the information in birth history.
In evaluating the DHS II datasets in terms of their quality; Marckwardt and Rutstein
(1996) mention that although DHS-II demographic data has better quality as
compared to earlier surveys “two of the most intractable problems, the age
displacement of women at the borders of eligibility and the displacement of birth
dates of children just outside the window of eligibility, remain unsolved”
(Marckwardt and Rutstein, 1996). They mention that in order to solve these two
problems, new solutions are improved such as using different interviewers to conduct
household and individual interviews and to make a change in field procedures which
will substantially cut the number of women excluded from interview because they
were mistakenly classified as having “slept away” from home. The level of
displacement of the birth dates of both mothers and children’s is greater in Sub-
Saharan Africa than other regions of the world. In addition, the overall assessment of
the DHS-II demographic data indicated that the side of gains is heavier than loses as
compared to DHS-I and WFS.
16
Macro International Inc. published the “occasional papers” series starting with “An
Evaluation of the Pakistan DHS Survey Based on the Reinterview Survey” (Curtis
and Arnold, 1994). With this study, the reliability of data collected in the Pakistan
DHS (1991) is evaluated through an analysis of the reinterview survey. Reinterview
survey is conducted 5 or 11 months after the Pakistan DHS with the 10% sample of
the women interviewed by using a much shorter version of the original questionnaire.
The reinterview survey focused on fertility and contraceptive use, with a reduced
number of questions on marriage and background characteristics of the woman and
her husband. The primary aim of the reinterview survey is to assess the reliability of
reporting of key variables in Pakistan DHS. Reinterview survey indicated that the
majority of the variables are not reported consistently. However the discrepancies
between the surveys are not mentioned as systematic. Especially the age and date
reporting between the two surveys are estimated inconvenient. In addition, although
the total number of living children information in two surveys was convenient; there
was a substantial displacement of births probably because of the attempts by the
interviewers to reduce their workloads. As the reporting of age and date effects many
rates calculated directly from the survey data, one of the important results was that
according to the reinterview survey the fertility rates and infant and child mortality
rates are underestimated.
The methodological reports published by Institute for Resource Development/Macro
Systems (IRD) carry valuable information in terms of evaluating the data quality of
the DHSs carried out (IRD 1990, Arnold 1990, Curtis 1995). By these reports the
design and implementation of DHS were examined to provide answers, explanations
which will be of benefit to survey researchers, particularly in developing countries.
These methodological reports consists the methods to evaluate the possible data
quality problems of DHS carried out in many countries. Reports show the methods
and results reached by the application of these methods in a comparative way. Some
selected countries were compared in terms of their data quality. Possible data quality
problems are mentioned and comparative studies were done in these studies.
17
Arnold (1990) in his section in An Assessment of DHS-I Data Quality focuses on the
data quality of the birth history data lists the possible sources of non-sampling errors
that can be seen at that section as; incompleteness of reporting of children’s birth
rates, displacement of children’s birth dates, age heaping, miscalculation of year of
birth and coverage problems of live births and their implications on the fertility rates.
He also makes comparison among the DHS datasets of selected countries from
different parts of the world.
In the same report a section is developed by Sullivan et.al. (1990) on assessment of
the quality of data used for the direct estimation of infant and child mortality in the
demographic and health surveys. They aimed to identify errors in data collection
which have occurred in a number of surveys and which may signal a need to modify
the DHS questionnaires or field procedures; to provide users of DHS data with
comprehensive, as well as survey-specific, information about the quality of the data
used to calculate childhood mortality rates. They mention the collection of DHS
mortality data, the completeness of the date of birth and age at death data, their
accuracy and impacts on estimates on mortality estimates. In addition, they focused
on the completeness of the event reporting which is mainly takes place in omitting
the death of the children. Like Arnold, they finish their section by making a
comparison among different countries’ DHS data quality.
On the other hand, Curtis (1995) in his study evaluated the data quality of second
phase DHSs focusing on the data used for the direct estimation of infant and child
mortality. He applied same tests with Jeremiah Sullivan and et al. for the DHS-II
data sets constructed in various countries. He found that two basic problems on
mortality data is the differential displacement of the date of birth of surviving and
dead children to the period prior to that covered by health section of the
questionnaire and the omission (or misplacement) of deaths that occurred more than
10 years before the survey.
Another study which is also noticeable is done by Brass (1996). In his paper
“Demographic Data Analysis in Less Developed Countries: 1946-1996” discusses
18
the chronologic development of demographic data analysis methods developed or
used for the less developed countries for a better estimate in fertility and mortality.
He focused on especially the indirect estimate techniques which help the
demographers to reach reliable results when they have inadequate or limited data. He
discuss the general techniques for the estimation of vital rates from censuses or
surveys like reverse survival, own children and P/F ratio. This is a brief history of the
developments in the demography in terms of demographic methods to analyze the
data collected from different sources.
The quality of the data collected by registration systems and censuses are still an
important interest area of the scientists. Data quality literature has many studies
focused on these types of datasets. Maged Ishak presented a paper discussing the
misreporting of age and underreporting of the death problem in vital registration
system in Egypt in his study (Ishak, 1999). As the mortality data, especially of the
childhood span, are known to be mostly affected data by under registration, he
presents some mathematical and regression techniques to model the under
registration behavior. His objective was to utilize the limited information together
with the available knowledge of underreporting in developing countries in order to
present a mathematical model that can reveal the age pattern of under registration. At
the end of his study, he combines the model with a regression model for infant
mortality to finally present an improved model that is free from the affects of both
the under registration and age misreporting problems to verify the model.
Age and sex data quality is studied in nearly all countries who conduct censuses.
Poston and friends conducted a comprehensive analysis of the quality of the age and
sex data of the country of the Republic of Korea (ROK) and its provinces. They
evaluated age and sex data gathered in the 1970 and 1995 Korean censuses
pertaining to the entire populations of the ROK, and also to its provinces in the two
time periods. They used indexes and methods described by Shryock and Siegel
(1976) and Arriaga and Associates (1994).In general although there is no clear
problem in reporting the age in both of the censuses, they realized that the quality of
age and sex data increased between 1975 and 1990 censuses. The important point at
19
this study is the Whipple index which focuses on the digits "0" and "5" did not seem
useful for the countries which the heaping is not commonly seen at these digits.
Therefore although there is a heaping at digit 5 Whipple index could not be able to
put forward this problem.
Another study is done by Bailey and Makannah (1996) which focused on the data
quality of age data held in African censuses during 1970s and 1980s. They work on
the single age distribution on the surveys by using Myers’s index, Bachi index,
Carrier index, and Ramachandran index. They find that age misreporting is more
obvious among females than males and among rural than urban residents. In addition
they found that literacy has a positive effect on the quality of data on age. They also
find that the countries where the age misreporting is relatively low, the coverage of
birth registration is better.
There are various papers in the literature focusing on the quality of surveys and
studies other than WFS or DHS type international surveys. US Census Bureau is
conducting “The Survey of Income and Program Participation” (SIPP) which
provides information at household level to analyze the economic situation of
households and persons in the United States. The roots of the program go to 1975
when the first Income Survey Development Program (ISDP), was initiated by the
Department of Health, Education, and Welfare to collect income and social related
data related with income. The techniques are developed to get better quality data at
these surveys and the structure of the program changed and the first questionnaires of
SIPP are applied at 1983. SIPP program published both the reports on the results of
the questionnaires and besides, reports on the data quality of the study to give light
about the quality of the survey carried out (Kalton, 1998; Killion, 2004 ).
The quality of the data on sensitive issues is focused by many academicians with
different methods. A study based on a validation survey conducted in Estonia in
April and May 1992 (Anderson et al, 1994). Sample of the study is the records from
the maternity hospital. With this study, the women who had abortion in 1991 were
20
asked about their health events by using a life history chart which links together on a
multicolumn grid such life-course events as marriages, divorces, pregnancies, births,
and abortions. It is estimated with this study that more than 80 percent of the women
reported their recent abortion. According to the study, this tool is helpful in
reconstructing pregnancy and abortion histories that date back many years.
The rates estimated from the surveys have also been questioned by the researchers
with further studies. Retherford, Mishra and Prakasam (2001) intended to put
forward the underlying reasons for the drastic changes in Uttar Pradesh state in India.
They used the data sets of National Family Health Surveys conducted in 1992-1993
(NFHS-1) and 1998-1999 (NFHS-2) and India’s Sample Registration System to
understand the possible reasons for the noticeable fertility decline at NFHS-2 as
compared to NFHS-1. They evaluated the survey datasets and the registration system
and; both assessed the distribution of age and sex by using Myers’ Index and focused
on the displacement of births which will mainly caused by interviewers which is
done purposively to lighten their workload in the field. They used to compare the
fertility rates with both birth history method and own children method. They find that
there is a great displacement and omission of births and misreporting of both
mothers’ and children’s ages in NFHS-2 as compared to NFHS-1 which leads the
great decrease in the fertility rates. In addition, they indicate that trend for the survey
years estimated by using the sample registration system is also affected by the
underregistration of births.
Retherford and Mishra’s interest on the quality of the fertility data of Indian National
Family Health Survey did not end with the Uttar Pradesh state. They also attempted
to evaluate the reasons for the discrepancies in National Family Health Surveys
1992-1993 (NFHS-1) and 1998-1999 (NFHS-2) and India’s Sample Registration
System (SRS) and to assess the true trend of fertility in all India and 16 major states
in India (Retherford and Mishra, 2001). They use both birth history method and won
children method to estimate TFR with the adjusted birth and age information. Age
misreporting and heaping is one of the major problems on the data in both the
NFHSs and SRS in India. They found that misreporting of women’s ages in the two
21
NFHS surveys does not have a large effect on the estimates of the TFR, for either
India or major states. However, misreporting does have a large effect on the
estimates of ASFRs. They calculated correction factors with the relation of General
Fertility rates estimated form NFHS to adjust the dataset of SRS to estimate fertility
rates.
Quality of the data used to estimate infant and child mortality is studied by different
researchers all over the world. A study in Egypt were done to determine levels and
possible trends and differentials in completeness of registration data of infant and
child deaths in Egypt, and to use these estimates to adjust reported levels of infant
and child mortality for the whole country (Becker et al., 1996). They used the two
survey data UN’s PAPCHILD-1991-1992 and DHS-1992 to estimate the
underreporting of the infant and child mortality in the vital registration system. They
selected the all deaths of children reported as born in five years before the date of
interview and checked these deaths in the registration system whether they are
recorded in the registration system. Their study indicates that only 64% of the cases
were notified by the registration system. Only 35% of the neonatal deaths and about
90% of the deaths above 1 year were notified. With these findings they calculated
correction factors for the registration data and found that not only the mortality rates
in registration system but the rates in the Surveys directly calculated from data are
about 21% level than the estimated levels.
In addition, Kingkade and Sawyer (2001) on their study focused on the problems of
the data in the former Soviet countries with regarding to the calculation of Infant
Mortality. They mention that the infant mortality rates estimated in the former Soviet
countries is vulnerable to the miscalculation and comparing them to other countries
indicators will lead to misunderstanding. Therefore they used the methodology to use
the mortality rates at ages 4-9 months is used to predict mortality in the earlier
months of infancy. With this methodology they adjusted national infant mortality
rates to improve their international comparability. They found that in most instances
the adjustment factors derived by this methodology are stable, or does not make
obvious trends in direction, except over very long time spans like decades. They
22
resulted that most East European countries have experienced sustained declines in
infant mortality after the 1990’s, while a number of former Soviet countries exhibit
stagnating tendencies. On the other hand, they put forward that Lithuania,
Azerbaijan, and the countries of former Soviet Central Asia are characterized by
clear declines in their adjusted IMRs.
One of the recent studies which examine the measurement of infant mortality is done
by Nadezdha and Redmond (2003). They studied the data of the countries of Central
and Eastern Europe and the Commonwealth of Independent States. They indicate
that the IMR is underestimated in most of the countries especially if the results of the
registrations system are used. They work on the strong and the weak aspects of the
surveys done in these countries which also give results on IMR. They resulted in that
although the surveys –especially DHS- have reasonable results on IMR; surveys are
rather blunt instruments, and that the confidence intervals that surround estimates
from these surveys are often large. It is suggested that to improve the collection of
official statistics on births and infant deaths in many countries across the region and
the effectiveness of surveys as a measurement tool further works needed to be done.
Loaiza (2004) in his study “Does data quality explain the differences in the current
global estimates for mortality and education?” discusses data quality as one of the
main issues associated with the found differences between the global estimates for
mortality and education produced by International Organizations. According to him
the statistics published by international organizations are not always coherent and
will be affected by the data quality problems. He focuses on the potential data
sources for under five mortality data and net enrolment/attendance ratios and their
problems in terms of reliability and quality. He suggest that vital statistics and data
obtained from administrative records are the most desirable and commonly used data
sources to monitor levels and trends or these two indicators. However, he puts
forward that these data sources are not complete or are not consistent with the
population to which they refer. He mentions the household survey data as an
alternative to improve estimates for these two indicators and to improve the routinely
systems of data collection.
23
Comparing the data quality estimations with other data sources is crucial for
understanding the extent of the problems. The Puerto Rico Fertility and Family
Planning Assessment (PRFFPA) were conducted in 1982 by the Puerto Rico. Warren
(1985) in his study examined the internal consistency of the marriage and birth
histories data from the individual questionnaire of the PRFFPA. In addition, he
compared his findings with external data sources (censuses, and vital registration
system and earlier studies). He evaluated the individual data set and found that there
is no clear evidence of omission of children or displacement of the births
Language barrier of the respondents to understand the questions well is also studied
in terms of the factors effecting data quality. Mc Govern intended to conduct a
quantitative assessment of data quality between non-English and English-speaking
households in the American Community Survey (ACS). With this research an
understanding of which language groups in the United States have the greatest
numbers of households with the lowest levels of English proficiency is discussed. In
addition, the research determined how these households are interviewed in the ACS,
and how complete the data collected from these households are. This research
addresses key questions about whether existing methods are resulting in the
collection of incomplete data in the ACS due to language barriers or not. It is
mentioned that linguistically isolated households had lower percentages of response
by mail than households speaking English only.
Malaysian Family Life Surveys (MFLS), fielded in Peninsular Malaysia in 1976 and
1988, is studied in terms of the quality of the retrospective answers (Beckett, et al.
2001). It is found that although some events may be forgotten over time, the answers
for the retrospective questions are consistent at the two studies. They suggested that
long-term retrospective histories provide nearly as good quality reports as provided
by short-term retrospective histories as indicated in the literature.
The effects of the data quality problems on the rates are also assessed in various
studies. Preston, Elo and Stewart (1998) in their paper, examined how age
24
misreporting can affect estimates of mortality at older ages. They checked the effects
of three patterns of age misreporting that reflect age overstatement, age
understatement, and symmetric age misreporting. They found that regardless of the
method employed, all three forms of age misreporting causes underestimates of
mortality at the oldest ages. In addition they indicated that “the age at which the
distortion begins varies according to the type of age misstatement that is present”.
Data quality of the longitudinal surveys is also studied by the researchers. By
comparing the responses of the same individuals for two rounds of the surveys in
Thailand, Knodel and Piampiti (1977) tried to measure the extent of the response
reliability for the questions asked both times. In addition to measuring individual
reliability they are also interested in the extent to which the aggregate results agree or
differ for the two rounds of the survey. It is found that, in general, reliability is much
higher for the behavioral variables than for the attitudinal ones. Moreover, the
highest reliability is found in responses of male household heads and married
women. They also found that “response reliability showed little relationship to
education level in the Longitudinal Study and urban respondents were characterized
by only slightly higher reliability than rural ones”.
To increase the level of data quality, new tools were tried and studies were done to
estimate the difference at the level of quality of data with or without using these
instruments. Becker and Sosa discussed on the results of the Costa Rica Family
Planning Survey in which a monthly calendar is used to record the events related
with for the five-year reference period before the survey for the recording of fertilityrelated
events (sexual unions, contraceptive use, pregnancies, and breastfeeding).
They showed that this is a better way to record these events as compared to the
traditional question formats. Moreover, it is indicated that with using calendar they
overcome the problem of inconsistencies and overlapping at the history of events
related with fertility (Becker and Sosa, 1992).
New technologies were also presented for the surveys which will be helpful in terms
of data quality. Using Personal Digital Assistants (PDA) at the surveys instead of
25
paper and pencil (PAPI) has introduced “computer aided personal interviewing
technology” (CAPI) to the surveys. The pros and cons of the CAPI surveys are
discussed by Bixby, Cespedes and Montero (2005) at the Population Association of
America Meeting in 2005. They reported the lessons learned from the CAPI survey
using PDA’s in Costa Rica. The cost effectiveness, the advantage of detecting the
data errors and making editing during the interview is mentioned as the most
powerful sides of using the PDA. Another important advantage is to handle the
complex questionnaires, including skip patterns, filters and rosters, which are
difficult to implement, and prone to error, when are on paper. On the other hand, the
important problem of using PDA is mentioned as the programming and preparing the
data entry program for a full-scale survey is mentioned as expensive because of using
highly qualified personnel.
Turkish literature on data quality of surveys and censuses are mainly focused on the
age reporting problems. In that manner, censuses are evaluated by various studies to
estimate the quality of age data. Demeny and Shorter (1968) discussed the overall
census quality and the age reporting problem of the censuses 1935 to 1965. They
made age corrections by Demeny-Shorter method for the census results. Following
this study Das Gupta (1975) and Ntozi (1978) also focused on age misreporting
problem in Turkish censuses and criticized the method of Demeny and Shorter and
offered new techniques to estimate and correct the age reporting problem in the
censuses. All of the studies attracted the attention of the age heaping problem for
both sexes, especially the women responses.
Mukherjee and Mukhopadhay (1988) and Güneş (1989) also studied on the methods
of correcting the age reporting problem of the censuses conducted in Turkey. Like
the previous studies, they found that there is a heaping problem for the ages ending
with “0” and “5”. Moreover, Albayrak in her thesis, studied on the correction of age
misreporting at the censuses of 1975, 1980 and 1985 (Albayrak, 1991). She indicated
that as the former studies on Turkish censuses, there is a coverage problem for the
censuses and although the age misreporting problem is corrected by different
methods, this coverage problem will still create the problem for this correction.
26
Another thesis on age misreporting in censuses is done by Tungul (1995). The
censuses between 1975 and 1990 are studied with this study. The results are more or
less the same with the earlier works; a decreasing level of the heaping problem for
the digits “0” and “5” which is more at women ages than men as the year of the
census is increasing. The coverage and response errors for the censuses assessed with
this thesis are also indicated by Tungul.
One of the latest studies was done by Yavuz and Coşkun (2002). They take the age
misreporting and age heaping problem into consider for all censuses conducted
during the republic period. They mentioned that, although there is a clear preference
for the digits “0” and “5”, as the overall digit preference is analyzed, even numbers
are preferred more than odd numbers. In addition the least preferred digits are found
as “1” and “9”.
Two studies concerning the data quality of the census results are done by Canpolat
(2002 and 2003). In the former study, she focused on the data quality of the age
reporting of the 10 selected provinces –according to the level of literacy- for 1990
and 2000 censuses. The results show that the level of errors decreased between the
1990 and 2000 censuses. She indicated that, the errors have relations with the
economical development and education levels; when the education level increases
the problem of age misreporting decreases (Canpolat, 2002). In her thesis of
expertise, she studied the age reporting problems and age correction techniques of
the censuses conducted in Turkey since 1935. She used Myers, Whipple and UN
Age-Sex Accuracy Index methods to estimate the level of age reporting problem and
by using the Arriaga, Carrier-Farrag, Karup-King-Newton, UN and Strong Moving
Averages techniques, smoothed the age distribution of the censuses (Canpolat, 2003).
In addition to the studies on censuses’ data quality –which are limited with the age
distribution problems- the results of the WFS and DHS were assessed in terms of
data quality. As a publication of the WFS scientific reports series Üner evaluated the
1978 Turkish Fertility Survey. The primary goal of the report is to assess the data
quality of the survey with the goal of comprehending the extent to which the
27
estimates of the demographic measures and variables obtained through the survey are
accurate and reliable (Üner, 1983). According to the results of the study, in general
the data from the birth history of the Turkish Fertility Survey was very good, but it is
mentioned that there will be a small level of omission and heaping problem in the
data. However, the problem of misreporting and omission is mentioned to affect the
single year estimates of the mortality. The most common error in the Turkish
Fertility Survey dataset is mentioned as the digit preference.
In his study “Fertility Trends in Turkey: 1978-1993” inside Fertility Trends,
Women’s Status, and Reproductive Expectations in Turkey Hancıoğlu (1997) aims to
revaluate the results of the Demographic Surveys in Turkey with focusing on the
fertility trends and levels. In addition, the data quality of birth history is examined to
understand the extent of the non-sampling errors of TDHS-93. He focused on the
data has direct effect on fertility rates. Although the standard DHS reports has a
section specifically prepared for overall data quality of the surveys, this study is
important for the discussion of the data quality of the variables used for the direct
estimation of the fertility. In addition to this study, although the main concern of the
study is not the data quality, Törüner (2001) in his master thesis worked on the
comparability of the national surveys conducted in Turkey and within his study he
also worked on the potential respondent and interviewer errors in these surveys.
The standard DHS main reports have always had a separate section (Appendix -D) to
bring the basic data quality estimations in to matter. Like other country reports, in
1993 and 1993 TDHS main report (MOH et al., 1993; HUIPS, 1999), standard tables
to show the basic response errors like heaping, omission and age and sex ratios of the
survey population. On the other hand, besides the tables, explanation of the data
quality tables was made at TDHS-2003 main report (Koç, 2004).
28
III. DATA AND METHODOLOGY
III.1. THE HISTORY OF THE DEMOGRAPHIC SURVEYS IN THE
WORLD
It is known that as compared to demographic surveys the censuses has a longer
history which goes back to ancient times in Egypt, Babylonia, China, Palestine and
Rome (Shyrock and Tauber, 1976). The importance of the demographic surveys for
developing policies for population programs is obvious. To understand the situation
is the first step to control and change it. Information on the population is needed to
develop policies and programs for different specific purposes. The main purpose of
the first demographic surveys was to decide the economic, labor and military
obligations of the population and limited with household heads, male population,
adults or taxpayers. Although censuses are subject to various types of errors it is
nearly universal and for nearly all national governments census has always seen as
the main source of information about their population.
On the other hand; to collect information on demographic events with surveys has a
history goes back to the Domesday survey in 1089. William the Conqueror made a
survey in England to learn the extent of the land and resources in order to foresee the
tax will be collected from this land. The information was collected and written into
two books however the study did not finished and has ended with the death of
William. This study will be accepted as the first survey conducted and published.
After that, various population surveys were conducted all over the world with
different purposes. The first study made using probabilistic sampling was the
investigation done by A. L. Bowley (1913) on the living conditions of the working
classes in Reading (England).
29
Raymond Pearl carried out one of the first demographic surveys in 1939 with 31,000
women in American hospitals. US Census Bureau conducted the Current Population
Survey (CPS) monthly since 1940. The 1960 Growth of American Families Study by
Whelpton, Arthur Campbell, and John Patterson; and the 1965, 1970 and 1975
National Fertility Surveys carried out by Charles F. Westoff and Norman B. Ryder of
Princeton University are the other important American surveys on demography. In
addition, the National Center for Health Statistics (NCHS) carried out six rounds of
the National Survey of Family Growth (NSFG) between 1973 and 2002. (Vaessen,
2008).
Like in U.S. other developed countries conducted more or less the same contented
surveys nearly same time period. Most of these surveys were designed to estimate
the labor force participation and were not specifically designed for demographic
purposes. With the Britain fertility study of David Glass and Eugene Grebenik in
1946, the fertility based demographic surveys were gained in speed and in 1960s in
Belgium, Canada, Greece, Hungary, The Netherlands, the United Kingdom, and the
Soviet Union fertility studies were done.
The 20’th century- especially after the second world war, the interest on the mortality
and fertility levels of the developing countries increased the worldwide research
projects. The high rates of population growth have been viewed as surpassing the
ability of countries to sustain socioeconomic development, reducing the resources
and causing major political instability. Instead of pro-natalist polices policy makers
and governments begin to support anti-natalist policies especially for the rapid
growing undeveloped and developing countries. The close relationship between the
eugenic theories should be kept in mind while evaluating the international
demographic survey programs. The theory focuses on the birth control for the
“systematic elimination of the so-called ‘undesirable’ biological traits and use of
selective breeding to ‘improve’ the characteristics of an organism or species” (Araujo
and Sommer, 2002). The aim to control the fertility of special groups needs better
knowledge and information on the fertility practices of the group. Therefore, both
financial and technical support by the developed ‘West’ for the demographic and
30
health surveys which were conducted on especially undeveloped and developing
areas of the world have been giving valuable information on the fertility levels and
movements at these areas.
Leaded by the Population Council, during the 50’s and 60’s, the surveys aimed to get
information about the needs, attitudes and level of information of people on
contraceptive use and limiting and/or spacing births were supported. Since 1970, 400
different surveys were conducted in 67 countries, most during 1960’s (Mauldin et.al.
1970). Just before the start of WFS program in 1970’s; independent demographic
surveys were conducted focusing on fertility in 15 European Countries. In addition,
at the Population Centre and the Department of Biostatistics of the University of
North Carolina to develop the survey techniques and to improve the methodology to
measure, analyze and evaluating the data on demographic processes the study
POPLAB (Program of Laboratories for Population Statistics) started in 1969.
Program continued all through 1970’s and 80’s in various developing countries
including Turkey. The data from surveys, censuses and registration systems were
evaluated and improvements for the quality of the data were proposed.
In 1990’s UNFPA supported the Fertility and Family Survey (FFS) Programme
followed by Generations and Gender Programme (GGP) (after year 2000) which
were conducted in United Nations Economic Commission for Europe (UNECE)
countries. The basic aim of FFS is to uncover the new fertility trends, marriage and
partnership relations and contraceptive behavior in Europe since 1960’s. The Center
for Population and Family Studies (CBGS) prepared the core questionnaire for FFS
in 1989-90 which were later shaped the final version of the core questionnaire by
FFS Informal Working Group. On the other hand, GGP aims to carry out
multidisciplinary and comparative studies to examine the family dynamics in
developed countries, and to study the relations between the generations, genders and
their effects on new demographic trends in UNECE countries. As making causal
deductions needs a lifetime data of the individual on demographic events, in a
prospective manner, GGP broadens the explanatory scope of the collected data.
Instead of applying the cross-sectional survey series, GGP aims to make panel
31
studies (waves) in 3 year periods to collect important variables like income and
opinions which is very difficult to be collected by retrospective questions. It is a
multi-country program covers relationships between generations also from the
viewpoint of the population above the reproductive ages, which allows analysts and
policy-makers to address the pertinent issues of population ageing in developed
countries (UNECE, 2008).The GGP is coordinated by Population Activities of
United Nations Economic Commission for Europe (UNECE PAU) and methodology
and questionnaires were developed by Max Planck Institute of Demographic
Research.
III.2. THE HISTORY OF THE DEMOGRAPHIC AND HEALTH SURVEYS
Most of the developing countries are lacking reliable sources for social, economic
and demographic information. As the vital registration system has problems in terms
of reliability and coverage; and censuses are very expensive and have structural
problems to obtain detailed information on various demographic issues, surveys are
the only main source for many interest areas for both the social scientists and the
policy makers. Sourcing from the interest to know and control the demographic and
epidemiologic transformations in developing countries, and the need to make
comparisons in international level; World Fertility Survey (WFS) is designed and
carried out in 41 countries during the period 1972-1984. Project was financially
supported by UNFPA, U.S. Agency for International Development (USAID), UK
Overseas Development Administration (UKODA), governments of France, Japan,
Netherlands and International Development Research Center of Canada; and
technical assistance was supplied by International Statistical Institute (ISI) and
UNFPA. There were two main model questionnaires: Household and Ever Married
(aged 15-49). However, in some countries never-married and male questionnaires
were also applied. All 41 internationally-funded developing country surveys
produced detailed national reports. In addition, the central office produced about 80
scientific reports, 50 multi-national comparative studies, 11 technical bulletins, 12
methodological studies, and at least 500 analytic projects based on WFS data. At the
32
very least, "these papers probably contain more information about the practical
methods of survey taking and interviewing in developing countries than has been
published in any previous enquiry" (Grebenik 1981: 25, cited by Blake 1983: 154).
Besides WFS, Contraceptive Prevalence Survey (CPS) designed specifically to
estimate main indicators of fertility and family planning was applied between the
years 1977-1985. The Centers for Disease Control (CDS) carried out the first CPS
which was lately funded by USAID and under the technical assistance of
Westinghouse Applied Health Systems (Measure Macro). The CPS were conducted
in 110 developing countries where the results were used directly for policy purposes.
The structure of these two important surveys became inadequate with the changing
interest areas of the demographers and the policy makers. Therefore, by merging the
characteristics of WFS and CPS and expanding the interest to mother and child
health issues the Demographic and Health Surveys (DHS) program is established at
The Institute for Resource Development, Inc. (IRD), a subsidiary of the
Westinghouse Electric Company in 1984.
Three phases of DHS are as follows: DHS I in 1984-1989, DHS II in 1988-1993, and
DHS III in 1992-1999. Since 1997 DHS changes its name to DHS+ to reflect a new
mandate under the MEASURE program. MEASURE DHS+ incorporates traditional
DHS features with expanded content on maternal and child health. Until now the
DHS+ program has provided technical assistance for more than 100 surveys in
Africa, Asia, the Near East, Latin America, and the Caribbean.
Demographic and Health Surveys (DHS) are nationally representative household
surveys with large sample sizes of between 5,000 and 30,000 households generally.
Although nationally representative sample of women ages 15-49 are interviewed and
basic criteria for interviews were this group of women, never-married in that range
also covered in many countries. In addition, along with female interviews, a subsample
of males was also interviewed in some countries.
33
The DHS combines the qualities of the WFS and the CPS and adds important
questions on maternal and child health and nutrition. The standard DHS survey
consists of a household questionnaire and a women's questionnaire. The household
questionnaire contains information on the following topics: household listing,
household characteristics, and nutritional status and anemia. The women's
questionnaire contains information on the following topics: background
characteristics, reproductive behaviour and intentions, contraception, antenatal,
delivery, and postpartum care, breastfeeding and nutrition, children's health, status of
women, AIDS and other sexually transmitted infections, husband's background, and
other topics like domestic violence and maternal mortality.
Fieldworks of the DHS were primarily conducted by national agencies. As well as
the country specific analysis, comparative analyses also were carried out by using
countries’ data set. Besides the dissemination networks of each country institution,
Measure DHS, makes distribution of hardcopy or electronic copy of the final reports
and other document from their web site. Data is disseminated through Macro
International and final reports are produced for most DHS surveys by the
collaboration with the national agencies. Comprehensive survey results are published
in the DHS Final Reports approximately 8-12 months after the completion of
fieldwork. Standard reports are approximately 200 pages in length and include, but
are not limited to, topics such as: household and respondent characteristics, fertility
and family planning, maternal and child health, nutrition, and HIV/AIDS. Besides
these country reports, there are comparative studies and analytical reports done on
some important issues held in the survey design.
34
III.3. THE HISTORY OF THE NATIONWIDE DEMOGRAPHIC SURVEYS
IN TURKEY
Turkey has a census history which goes back to Ottoman Empire period. First census
attempts were done between 1326 and 1360 to follow up the military potential and
the extent of the lands. Moreover, the first successful census was done in 1831 by
which the male population was counted to estimate the potential military population
and new tax sources (TURKSTAT, 2008). After the establishment of the Turkish
Republic in 1923, first census was taken at 1927; starting with 1935 census until
1990; every 5 year census was held by TURKSTAT. In 1997, an enumeration was
done specifically for the decision of the population give vote in the elections. The
latest census was taken in 2000 and after that an effort was made to solve the
problems in vital registration system. With the changes in population law in 2006,
the new population registration system, which will be the main data source of
population censuses, was established in the country hand in hand with MERNIS
registration system. The ‘National Address Database’ that covers all addresses in the
boundaries of the country was established and Turkish citizens living in these
addresses were registered according to the ‘Turkish Republic Identification
Numbers’, whereas the foreigners were registered to the system with their passport
numbers. According to the ‘Address Based Population Registration System’ the
population of Turkey is calculated as 70,586,256 for December 31, 2007.
After the Second World War, parallel to the interest on making plans for health
purposes have increased the awareness in demography in the world and Turkey as
well. In a situation where no reliable information was brought by about nearly all
demographic rates; surveys seemed as a solution to understand the demographic
situation of Turkey. The nationwide health and demographic surveys in Turkey have
always been designed and carried out as a part of the worldwide demographic
surveys. Nearly for all nationwide surveys, the timing were planned to bring
demographic information for intercensal periods.
35
The first nationwide demographic study was the 1963 Turkey Demographic Survey
carried out by School of Public Health (SPH) with the financial support of
Population Council. Two types of questionnaires were used; household and ever
married women, and 9701 households and approximately 8000 eligible women
interviews were done. It was a round of KAP surveys carried out at the first half of
the 60’s in 20 developing countries. It was a policy oriented study and the results
were directly used to create a population program for Turkey.
During 1965 and 1968 SPH conducted dual-record survey which was organized as
collecting the data from two independent sources; registration system and household
survey data. The resident registers were collecting data by monthly visits to the
selected households, whereas SPH field staff was visiting the same households with a
6 month interval. After the success of this study, State Institute of Statistics (SIS)
conducted 1974-75 survey which has the same structure with the 1965-68 study.
However, the results of the study were not successful as compared with the previous
one and criticized as being in low quality.
III.3.1. Demographic Surveys carried out by HUIPS
The history of the nationwide demographic and health surveys is directly related with
the history of HUIPS. With the collaboration of Ministry of Health and other
governmental organizations and with the financial support and technical assistance of
the various international foundations and companies, HUIPS carried out 8 national
demographic and health surveys and other national and small scale surveys on
demographic subjects during the 40 year.
After the establishment of the institute in 1967, one year later, Institute carried out
Survey on Family Structure and Population Problems in Turkey (1968). It is aimed to
examine the main demographic structure and family formation and relations of
Turkey and their socio-economic and cultural determinants as well as family
planning issues (Timur, 1972). No survey report was published after the survey
36
whereas two books of Çavdar and others (1971) and Timur (1972) published by
Hacettepe University where the information on the survey can be found.
In 1973, the aim to conduct nationwide demographic surveys within the censuses
was brought to life with the second demographic survey carried by HUIPS. With the
financial aid of UNFPA, 1973 Turkish Population Structure and Population Problems
Survey demographic situation and trends were tried to put on. After the survey, The
Second Demography Conference was organized in Çeşme (İzmir) in 1975 where the
findings of the survey were discussed. However the main report of the survey was
able to be published in 1978 (HUIPS, 1978) at the time of the third nationwide
demographic survey: Turkey Fertility Survey. As a part of the World Fertility survey
programme, HUIPS carried out the survey with the financial aid of the UNFPA. In
addition to household and ever married questionnaire, husband questionnaire was
planned to be applied at the field, however, by extending the household questionnaire
with additional questions, husband questionnaire was not used.
The fourth national demographic survey was carried out in 1983 as a part of the CPS
programme. The focus of the 1983 Turkish Fertility, Contraceptive Prevalence and
Family Health Status Survey was to collect information on fertility, mortality,
contraceptive use and other health issues of women and child. Hacettepe Foundation
and Westinghouse Overseas Corporation Public Applied Systems and the
Demographic Data for Development Project funded the survey. Data was analyzed
by HUIPS and in 1987 the main report was published.
After one year publishing the main report of the 1983 survey HUIPS handled the
1988 Turkey Population and the Health Survey funded by USAID-Center for Disease
Control. The information collected by the survey did not changed much, it covered
the basic demographic indicators of fertility and mortality and contraceptive use and
health services. After 1968, this is the first survey by which the Husband
Questionnaire was used for half of the selected households. The report writing took
one year and the main report was published in 1989 (HUIPS, 1989).
37
III.3.2. Turkey Demographic and Health Surveys
In Turkey, three DHS (Phase 3) were carried out by HUIPS in 1993 (TDHS-93),
1998 (TDHS-98) and 2003 (TDHS-2003) (fieldwork of the latest leg hang down to
first months of 2004) respectively. The importance of these DHS and the previous
quinquennial surveys is apparent for Turkey as being the only nationwide surveys
which provide information on mortality, fertility and migration in general; infant and
child mortality, mother and child health, social and economic composition of the
households, contraceptive usage, migration, vaccination of the children and the
anthropometric characteristics of the mothers and their children aged under five, in
particular.
TDHS-1993
Main funding sources of THDS-93 were Macro International Inc. and State U.S
Agency for International Development (USAID).The fieldwork was carried out
August to October in 1993. Sample size of the survey was 8619 households and 5257
women. Two different questionnaires derived from DHS questionnaire format are
used. After the household, eligible ever-married women aged 12-49 were
interviewed.
TDHS-1998
The TDHS-98 was conducted through an agreement with Macro International Inc.
under the auspices of the MEASURE DHS+ project supported by the United States
Agency for International Development. It is the second demographic survey carried
out in collaboration with Macro International Inc. In addition, the contributions of
the United Nations Population Fund also were critical in realization of the survey in
its scope. Survey was implemented between August and November 1998 and 8,059
household, 8,576 women and 1,971 husband interviews were completed.
38
TDHS-2003
The TDHS-2003 was implemented by HUIPS, in collaboration with the General
Directorate of Mother and Child Health and Family Planning of the Ministry of
Health. Financial support for the TDHS-2003 was mainly provided through the
national budget as a three-year advanced project in the investment program of the
State Planning Organization. Moreover, the TDHS-2003 was supported for the first
time as a project in the frame of the European Union “Turkey Reproductive Health
Program”, implemented by the General Directorate of Mother and Child Health and
Family Planning of the Ministry of Health. The TDHS-2003 is the last leg of the 8th
national sample surveys carried out by HUIPS. 10,836 household and 8,075 ever
married women questionnaires were applied during the field study started in
December 2003 to May 2004.
Demographic and health surveys mainly have modules on family planning, maternal
and child health, child survival, HIV/AIDS/sexually transmitted infections (STIs),
and reproductive health. On the other hand, country specific modules/questions can
be added to the core questionnaire. The contents of the questionnaires used in three
TDHSs all were based on the International MEASURE/DHS+ survey project model
questionnaires and the questionnaires that had been employed in previous Turkey
population and health surveys.
39
Table III.1. Sample Size and Completed Interviews at Demographic and Health
Surveys, Turkey 1993, 1998, 2003
Demographic and Health
Surveys, Turkey
1993 1998 2003
Household
number of households selected 10631 9970 13049
number of household questionnaires completed 8619 8059 10836
Ever married women
number of eligible women 6862 9468 8447
number of women questionnaires completed 6519 8576 8075
Male
number of eligible males - 3043 -
number of male questionnaires completed - 1971 -
III.4. DATA SOURCES
In this study 1993, 1998 and 2003 Turkey Demographic and Health Survey data sets
are going to be used in terms of evaluating the data quality specifically on the
variables used to estimate fertility and early age mortality rates.
In this study, although the main focus is on the individual questionnaire and
especially on the quality of birth history section, the quality of the data used to
decide the eligibility of the ever married women is also going to be discussed. As the
selection of the ever married women starts from the age and sleeping last night (de
facto) information of the female members of the household is crucial for selecting the
eligible women from the household list.
In DHS surveys, a birth history section on ever-married woman questionnaires is
used by which information about all the live births of the woman were purposed to
40
be obtained. Using birth history module helps for both the interviewer and the
respondent. If the woman is at the end of her fertility period (e.g. in the age group
45-49), then she will have problems to remember the time of the births. Getting the
birth history of a woman in a chronological order will decrease the possible recall
errors and the literature stresses the handy characteristic of the birth history.
In TDHSs birth history module is placed at the “Section 2A: Reproduction”. At that
section, questions from 201 to 210 (Q208, q209 and 210 are the filter questions -they
are not directly asked to the women, but filled by the interviewer from the
information from the answers of the respondent-) were asked to all the eligible
women, by which the total number and the sex of the live births of the respondent are
going to be written down, which will help the interviewer at the birth history section.
Questions from 201 to 210 are:
201: Now I would like to ask about all the births you have had during your life. Have you
ever given birth?
202: Do you have any sons or daughters to whom you have given birth who are living with
you?
203: How many sons live with you? And how many daughters live with you? IF NONE,
RECORD “00”.
204: Do you have any sons or daughters to whom you have given birth who are alive but do
not live with you?
205: How many sons are alive but do not live with you? How many daughters are alive but
do not live with you? IF NONE, RECORD “00”.
206: Have you ever given birth to a boy or a girl who was born alive but died later? IF NO,
PROBE BEFORE RECORDING: Any baby who cried or showed signs of life but only
survived a few hours or days?
207: In all, how many boys have died? And how many girls have died? IF NONE, RECORD
“00”.
If woman has at least one or more live births, interviewer completes the birth history
section by asking questions about all the live births of the women one by one. For
every live birth, the questions above were asked:
41
Q212 : What name was given to your (first/next) baby? WRITE "BABY IF THE BABY DIED
BEFORE A NAME GIVEN.
Q213: RECORD SINGLE OR MULTIPLE BIRTH STATUS.
Q214: Is ….. a boy or a girl?
Q215: In what month and year..... born? PROBE: In what season was s/he born? NOTE:
FOR ALL CHILDREN, THE YEAR OF BIRTH; FOR CHILDREN BORN AFTER 1998, THE
MONTH OF THE YEAR OF BIRTH MUST BE DETERMINED.
Q216: Is …… still alive?
Q217: How old was .... at his/her last birthday? RECORD AGE IN COMPLETED YEARS.
MAKE CALCULATIONS FOR CONSISTENCY.
Q218: Is...... living with you
Q218A: RECORD THE LINE NUMBER OF CHILD IN THE HH LIST. IF S/HE WASN'T
RECORDED IN HH LIST, RECORD "00".
Q219: IF DEAD: How old was ...... when he/she died? IF “1” YR., PROBE: How many
months old was .....? RECORD DAYS IF LESS THAN 1 MONTH, MONTHS IF LESS THAN
TWO YEARS OR YEARS OTHERWISE.
Q221: Were there any other live births between (NAME OF PREVIOUS BIRTH) and .........?
In addition to the information collected with the questions above, interviewers are
responsible to complete the calendar module –which covers the 5 year period
preceding the survey-. Although the extent of the information needed to complete
calendar module, the basic aim of it is to place pregnancies (whether completed with
a delivery or not), contraceptives used- reason not to continue and to determine the
time that the women lived married during 5 year period preceding survey.
42
III.5. METHODOLOGY
III.5.1. The Assessment of Data Used to Determine Eligibility for the Individual
Interview
While studying the data quality of the birth history of the TDHSs, the assessment
should be started with the household questionnaire. Since, in all surveys, age is the
principal criterion used to determine eligibility for the women’s individual interview,
it is important to assess the quality of age reporting in connection with the household
interview (Marckwardt and Rutstein. 1996)The quality of the household data is
highly interrelated with the rates and ratios estimated from the individual
questionnaire where the birth history section is placed. The age distribution of the
household, the exclusions at the household with the sleeping away responses should
be studied to understand the overall data quality of the TDHSs which are highly
integrated with each other.
DHS Programme developed 2 types of questionnaires, Household and Individual. All
countries used DHS structure and questionnaires applied household questionnaire to
take the characteristics of the households and decide the members selected for
individual interview. Depending on the countries’ needs and socio-demographic
characteristics, while some countries collected information from all women at
reproduction ages (commonly 15-49 ages), in some countries only ever married
women at reproductive ages were interviewed. In TDHS-1993 and TDHS-2003 only
household and ever married women questionnaires were applied. Whereas, in TDHS-
1998, in addition to these questionnaires, never married women (aged 15-49) and
husband questionnaires (for half of the ever married women’s husbands) were used.
While the household questionnaire is applied for all households selected by the
sampling procedure, individual questionnaires are used for the members who are
selected to be interviewed after the eligibility criteria are supplied. The basic
criterion for eligibility is “the age”. For the household selected by the sampling
43
procedure, first of all, a Household Questionnaire is applied where all the household
members were listed. Age and sex of the members listed at the household list are
used to select the eligible persons to make individual interview.
III.5.1.1. The Assessment of the Data at Household Interview
III.5.1.1.1. Household Interview Results
The households selected after the sampling procedure -described in DHS Sampling
manual (IRD, 1987)- were aimed to be interviewed. In order not to change the
sampling frame, replacement is not allowed. It is known that replacement affects the
selection possibility given to each dwelling. The data quality of the data is directly
affected with the response rates of the selected household interviews. Sample
weights are calculated after each survey to overcome the problem of response errors.
If the response rates are higher, the weights calculated to reflect the response rates to
the data will be higher to compensate this problem. The household result codes are
evaluated by region and type of place of residence at this study.
The response rates are estimated for each survey as an indicator of the success of the
survey sampling. The response rates are estimated by dividing the number of
completed household questionnaires by the sum of completed, Household present but
no respondent, Postponed, Refused and Dwelling not found results. The formula to
estimate the Household Response Rate is below.
*100
C HP P R DNF
HRR C
+ + + +
=
44
III.5.1.1.2. The Quality of Age Reporting in Household Questionnaire
The one single variable included without exception in every demographic data
collection exercise is ‘age’, and it is thus the most widely studied and examined
demographic variable. Eligibility for inclusion in the survey of women age 15-49
rely on the age given in the household survey. Eligibility of children for questions on
health depends on the birth dates given in the birth histories. Both the numerators and
the denominators of age-specific fertility rates, infant mortality rates, and other rates
depend on reported age. In addition, the quality of the reports of ages and dates
reflects on the quality of other information in the surveys. (Pullum, 2006). Although
its importance and in spite of generations of research, age continues to be a variable
on which it is difficult, if not possible, to get good quality data in most populations of
the developing world (Chidambaram, et.al. 1984). Although the basic information on
a population is given by age, various studies indicated that the good quality of age
information is still not very possible in some developing areas of the world. Age
distribution of a population gives valuable current and historical information to the
researchers. The quality of the age data of the censuses and surveys gives basic
knowledge on the quality of the other information collected. Two basic forms of
misreporting of ages are “heaping” (digit preference) and “shifting”. In order to
estimate the possible errors on the distribution of age and sex, various techniques and
methods were developed and used. The most common ones are; Myers, Whipple and
Bachi Indexes and UN’s Age Sex Ratio Technique. At this study TDHS-1998,
TDHS-1998 and TDHS-2003 household data is evaluated in terms of age and sex
data quality by using these techniques.
Whipple Index
Whipple Index gives information about the digit preference for the ages ending with
“0” and “5”. Commonly for this index only the ages between 23 and 62 are taken
into consider because outside these ages the range of shifting and other problems
often tend to confuse the normal pattern of heaping (Newell 1988, Shryock and
45
Taeuber 1976). Within this range, the percent of the people age mentioned ending
with “0” or “5”at total population within this age group multiplied by 5.
The Formula for Whipple index is:
W =
* 5 *100
(23 62)
(23 62)
-
-
P
Py
where,
W : Whipple Index
y(23-62) P : population in ages 23-62 mentioned as aged years ending with “0” or
“5”,
(23-62) P : total population in ages 23-62.
Index takes value between 100 and 500, and while the index score 100 indicates that
there is no accumulation at the digits “0” and “5”, score 500 mentions the fact that all
the ages are mentioned ending with either “0” or “5”. Table III.5.1.1.1. indicates the
score to evaluate the age quality with Whipple Index (Newell,1988):
Table III.5.1.1.2.1. Whipple Index Score for Estimating Reliability of Age Data
Quality Whipple Index
Highly accurate under 105
Fairly accurate 105-110
Approximate 110-125
Rough 125-175
Very rough above 175
46
Myers Blended Index
This index shows the preferences or dislikes for each of the ten digits, from 0 to 9.
As the technique gives the preference or avoidance results for each digit, it gives
more detailed results as compared to Whipple Index. Technique takes successive
sums of numbers recorded at ages ending in each of the digits. In theory, all the
digits should have the 10 percent share; Myers Index puts on the deviance from this
“equal” distribution. The total summary of the digit preference is the sum of all digit
deviance in absolute values. The theoretical range of Myers index is between “0” to
“90”. If there is no digit preference in other words, all the digits were preferred
equally, and then the index is “0”. On the other hand if only one digit is preferred for
all population then the index value equals to “90” (Shryock and Siegel, 1976). If the
Myers index score is below “10”, the digit preference is very low and acceptable. If
the score is between “10” and “20” the digit preference is medium level and if the
score is above 20 the digit preference is mentioned as high.
Bachi Index
Another method for estimating the digit preference is the Bachi index. Although
Bachi index has some theoretical advances as compared to Whipple and Myers
indexes, as the calculation procedures are more laborious and the in general the
extent of heaping differ little from other indexes (Shryock and Siegel, 1976). Bachi
method gives estimations for each digits between the ages 23 and 77. The index
score is the same like Myers index and gives the absolute total deviations from 10
percent equal distribution. Like Myers, the theoretical range for Bachi index is from
0 to 90. If there is an equal digit distribution at the population, the index is 0, if all
everyone prefer one certain digit then the index score is 90. The results obtained by
Bachi index is similar to with the results obtained by Myers index.
47
United Nations Age-Sex Accuracy Index
In order to summarize the accuracy of a population in terms of age and sex
distribution, UN developed the Age-Sex Accuracy Index in the early 1950’s (UN,
1952, 1955). This index is also called as “Joint Score Index” and the mostly used
method to estimate the 5 year grouped age data by sex. It is generally used to
compare the age and sex data of different countries from all over world and to
evaluate their data quality. For every age group age specific sex ratios are estimated
(Sex ratio is calculated by dividing the number of male population by female
population multiplied by 100). After calculating sex ratios for each possible age
groups; age ratios were estimated by dividing the sex ratios of an age group by the
previous and later age group. Three indexes were calculated from this method:
(1) The index of sex-ratio score (SRS): The mean difference between sex ratios for
the successive age groups, averaged irrespective of sign.
(2) The index of age-ratio score (ARS): The mean deviation of the age ratios from
100 percent, also irrespective of sign.ASR is calculated for both males (ASRM) and
females(ASRF) seperately.
(3) The joint score (JS) or age-sex accuracy index: It is based on empirical
relationships between the sex-ratio scores and the age-ratio scores and calculated
with the formula:
JS = 3xSRS + ARSM + ARSF
The results of the joint score can be evaluated as follows:
(a) accurate: if the joint score index is under 20,
(b) inaccurate if the joint score index is between 20 and 40, and
(c) highly inaccurate if the index value is over 40.
Although the index can be effected by war, migration, epidemics etc. As it gives
results on the accuracy of both age and sex, it is commonly used and can be
48
mentioned as more effective than Myers and Whiplle Indexes. (Arriaga, 1994; UN,
1955).
Age information of the respondent is going to be studied in the household
questionnaire. Digit preference is going to be studied by using Myers Blended Index
and Whipple Index. In addition, age-sex accuracy index is going to be used to
evaluate the age and sex distribution at the household level. While examining the
quality of the age data at household survey, some important characteristics (age, sex,
relationship to the household, education level, etc) of the respondent with whom the
questionnaire is filled are going to be discussed.
III.5.1.1.3. Boundary Effects
The age, sex and current marital status information collected from household
interview constructs the base for the individual interviews. In 1993 and 2003 TDHS
fieldwork, after completing the household interview, if there was eligible woman -
aged 15 to 49 and ever married- ever married woman questionnaire were applied. On
the other hand, besides the 15-49 aged ever married woman questionnaire, never
married women questionnaire at the same ages and for the half of the eligible ever
married women’s husbands, husband questionnaire was applied. Within these three
criterions to ascertain eligibility the most critical and easily manipulated criteria by
the interviewers is “age”. Interviewer has the chance to lessen the workload by
making small arrangements on the age. To have a standard application for the
possible situations which the interviewer may face with while collecting the data,
during the TDHS training period the ways to overcome the problems on collecting
age were studied carefully and all the possible situations were discussed and standard
rules were put for common situations. Moreover, also during the training the
supervisors of the field team were told to follow some interviews or revisit some
households to check the information collected by the interviewers. As expected, to
revisit all the households for this purpose seems unnecessary and very difficult.
49
As mentioned above the interviewer who completes both the household and the
individual questionnaires may keep the eligible woman aged closer to the eligible
ages out from the eligibility criterions. This exclusion will be more likely fotr the
women who are unsure about their exact ages. In addition to the interviewers
purposive exclusion, because of individual reasons, the respondent for the household
interview may misinform the interviewer about the age of some members of the
household (Rutstein and Bicego, 1990). In both situations, the number of eligible
women will decrease and depending on the size of this exclusion, there will be a
possibility for miscalculation of indicators which uses the information of number of
women in eligible ages.
The previous studies show that this exclusion is more possible for the ages on older
ages. The level of education among women decreases with age and the women at
older ages may have problem at remembering their exact age. At that point, during
the training the interviewers were told to probe to get an average age. However, the
interviewer will not probe and will round the age and left an eligible woman outside
the survey. On the other hand, as the women at older ages are completed their
reproductive period and have all the pregnancies and births, an interviewer who
thinks to escape from the workload to deal with all the pregnancies and live births of
such women will push women out of the eligible age ranges.
The calculation of the sex ratios at the five year groups for the first and last eligible
age groups (15-19 and 45-49) and the age groups just before and after the eligible
ages (10-14 and 50-54) will reflect the exclusion of eligible women before and after
the eligible ages. If there is a systematic exclusion at the data then the sex ratios at
the first and last eligible ages will be higher than the last and first non-eligible age
groups because of the lower percent of woman at the eligible ages.
At this section three types of boundary effects indices can be calculated: Upper
Boundary and Lower Boundary Effect Indices and Total Boundary Effect Index. Age
and sex ratio of the boundary ages for eligibility are used to estimate these indices.
Lower Boundary Effect Index formula is:
50
LBE= (AR15-19-AR10-14) – (SR15-19-SR10-14)
Whereas Upper Boundary Effect Index formula is:
UBE= (AR45-49-AR50-54) – (SR45-49-SR50-54)
On the other hand, after the calculation of Lower and Upper Boundary Effect indices,
these two indices results were used to estimate the total Boundary Effect with the
formula:
TBE = |LBE| + |UBE|
As the total boundary effect index gives absolute value it only gives the total
distortion but not the direction. The total boundary effect index will be evaluated
within these ranges:
00-24 : Negligible
25-49 : Low
50-99 : Moderate
100+ : High.
III.5.1.1.4. The Household Residency
TDHS household questionnaire is used to select the eligible person for individual
interview. Besides the criterion age and sex, slept last night information is used for
eligibility. The standard DHS questionnaire collects information for each person
indicated as a member of the household and the visitors slept last night at the house.
Like a usual member if a visitor slept last night at the household selected he/she is
also selected for the individual interview. The de facto selection process is used at
TDHS to select the all women and to avoid the possibility of selection of a woman
twice. Like the other two criterions to be selected, “slept last night” is also crucial for
the data quality and should be studied.
51
TDHS-1993, 1998 and 2003 household data is going to be studied in terms of
household residency which can be identified at household list. In theory, the percent
of usual residents away from home should be equal to the visitors slept last night.
If there is a large difference, this will be a sign of exclusion of usual resident eligible
women from the household. Small difference will be accepted and will be a
reflection of the women staying at institutional places (like, hospital, dormitory, etc.).
The interviewers might record the woman not at home during the visit as “not slept
last night” and want to escape from the burden of interview with them. If this
intentional exclusion is large, then the estimations from the data will be biased.
III.5.1.2. The Assessment of Age Data at Individual Questionnaire
The woman questionnaire in TDHS starts with collecting birth month and year in
addition to the completed age information. Although the age information is taken at
household list, the information at the household questionnaire may be different from
the information got directly from woman. In some situations the women who are in
very close ages to the eligible ages may be mentioned eligible for individual
questionnaire although they were not. With the questions at the beginning of woman
questionnaire such kind of an error may be noticed and by cancelling this interview
and the interviewer will turn back to household questionnaire to change the age
information for this woman.
The age information gathered from woman is the basis for TFR and thus very
valuable. To reach the correct age information, during the training, interviewers were
emphasized to get the exact age. When the women do not know her exact birth date
the interviewer are trained to probe by using some important dates and/or seasons to
get the exact month and year information. If the women have no idea about the birth
year and month the interviewer is told to guess her age. “Age and year table” is used
to check the year-age consistency. While calculating the age information to year or
vice versa the interviewer may make mistake, however, at this table according to the
52
information whether the respondent celebrated her birthday at the survey year, it is
better to check the information from the table instead of calculating from mind.
Question 103 and 104 in TDHS-1993, and 105 and 106 in TDHS-1998 and TDHS-
2003 are the questions by which the birth month and year and the age information of
the respondent is collected. At this study the information of the women questionnaire
is also assessed.
III.5.1.2.1. Digit Preference
The Myers, Whipple and Bachi Indexes which are used to estimate the overall data
quality of the household member’s age is also going to be used to understand the
extent of the digit preference at woman questionnaire. In addition to these
calculations, five year age group distortions are going to be computed to understand
the total picture of the age distribution problem.
III.5.1.2.2. Imputation at the Age Data
During the field survey, it is highly recommended to get information directly from
the respondent not to impute the information during the data entry process. It is
important to get complete information directly from the women in terms of data
quality and the estimations done based on the women’s age. At this part of the study
the amount of the imputed age information of the women is going to be discussed.
III.5.2. The Assessment of the Quality of the Birth History Data
Retrospective questions have potential problems in terms of remembering the date of
the interest event and even the event itself. Woman may have problems in
remembering all the information about all her live births. Therefore, interviewer is
responsible for probing the women to remember the exact date of the child is born
53
and -if dead- death. Interviewer is responsible to catch any types of internal
consistency inside the questionnaire and probe as much as possible to overcome from
the inconsistency. Interviewer will not catch some of the inconsistencies at the
fieldwork. However, computer programs designed specifically on data entry catches
all terms of inconsistent information in the questionnaires. At that point imputation
gains importance. If there is inconsistent information about the date of the birth,
imputation is made. If there is an imputation on the date of the birth of the child, then
a flag is placed by which the data user understand that whether this information is
imputed or not. Imputation will be done in the field but it is impossible to know its
extent and/or it may be imputed during the editing process at the data entry. The
further studies mention that if the number of the births increases women has the
possibility of recall errors, and this will increase for the women who have lowest
education. In this study the socio-demographic characteristics of the women and the
data quality is also going to be studied.
All methods to assess the quality of the birth history data are going to be applied for
1993, 1998 and 2003 TDHS datasets. Hence, a comparison between the three surveys
will be done and the impacts of the possible quality problems on fertility rates can be
evaluated.
III.5.2.1. The Quality of Birth Related Data
III.5.2.1.1. Completeness of the information of Birth Dates of the Children
To evaluate the data quality of the data on the children’s birth dates generally
contains the completeness of the information. The completeness of the information
on birth dates of the children is related with the quality of the estimations done based
on this information. The data is directly collected by the mother of the children and
the questionnaire aims to collect both the birth month and birth year of each child.
While collecting the data from the women about the birth date of their children,
women may have difficulty in remembering the exact month and year of birth of
54
their children. For these situations “Don’t know” answers are recorded as “98” and
“9998” for the month and year respectively. The completeness of the birth dates and
the level of imputation is evaluated in the categories: “No imputation” “Month and
age reported -year imputed” “Year and age reported - month imputed” “Year
reported – age and month imputed” “Age reported – year and month imputed”
“Month reported – age and year imputed” and “All imputed”.
III.5.2.1.2. The Displacement of Children’s Birth Dates
The problem of carrying the birth dates of the children out of the five year period
gives the interviewer to escape from the workload of asking additional questions in
the next sections of the questionnaire about the children under five. DHS Program
has emphasized the collection of accurate data on demographic events and indicators.
During its three phases over the last decade, the scope of data collection has
increased tremendously. Beginning with preexisting substantial questionnaire based
on the WFS that included a full birth history, a contraceptive history and many other
topics, a great expansion of the survey instruments has occurred. (Marckwardt and
Rutstein,1996). The additions to the questionnaires consist numerous questions for
young children, anthropometric measures of the mother and the children, work and
occupation of women and the husband, reproductive health questions, HIV/AIDS etc.
If the woman gave birth within 5 years preceding the survey, then many questions
should be asked to this woman about the health, breastfeeding and immunization of
the child. Previous studies indicates that the interviewers who want to escape from
the workload for the child under five interviewers may change the birth date and
carry child to the age 5 or more. This displacement effects all the estimations created
for the under five children aged under five. The magnitude of the displacement will
have direct effect on the indicators estimated. If the interviewers move the birth date
in few month, this will make a small problem, however, if the displacement is for a
whole year then the data will be effected.
55
It is difficult to evaluate the extent of the displacement; however, the distribution of
living children by their ages will show the extent of the displacement problem. If
there is a clear displacement to the age 5 and more the percent of the children at this
ages in the birth history should be clearly higher whereas the percent of the children
at age 4 should be clearly low. In TDHS 2003, different from TDHS-98 and TDHS-
93, in order to calculate an average time spent for each section, the hour and minute
that the interviewer started to the section and finishes is recorded. The problem is
going to be discussed in terms of the length of the section about the children under
five. In addition the duration of the interviewer in the field, and the comparison of
interviewers with other interviewers is done in terms of evaluating the displacement
of the children to the age 5 and more.
Age Heaping
Another problem at the birth history data will be the age heaping. Digit preference is
one of the common problems for censuses and surveys nearly all over the world for
nearly all age groups. Especially in the underdeveloped and developing countries,
typically for the ages ending 0 and 5, a common preference is seen. Theere are
various studies which indicates that the age heaping increases if the education level
decreases and the age increases. In addition, the year of the birth may be
miscalculated by either the respondent herself or the interviewer and this will make a
data quality problem and should be evaluated. At this study the age ratios for the
children at the birth history is evaluated to understand the level of age heaping
problem for the birth history data. Estimations are done for the age and education
level of the mother and the time period of the interviewer at the field.
Miscalculation of Year of Birth
If the respondent doesn’t know the birth month of the children, either the respondent
or they may calculate the age of the children by easily subtracting the year of birth
from the year of interview. If the month of birth is not known but the age of the
56
respondent is not equal to the result of the subtraction of year of interview and year
of birth then the age is imputed. The level that the imputation is done is important.
III.5.2.1.3. Coverage of Live Births
Respondents will forget to declare their children who have died, who are not living at
home or very young. A check from the household survey will catch the misinformed
number of children in individual questionnaire and household list. A comparison will
be done for this reason.
Median age at birth and sex ratios at birth should be estimated to give an idea of the
exclusion of the births.
Covering all the live births is one of the most crucial issues in TDHSs, not only the
ones 5 year preceding the survey, all the births should be recorded down by the
interviewer. Not only the ones living with the respondent; the ones not living at
home, who have died and who are very young are specifically asked to the
respondent to fully cover all live births. With this study, the possible coverage errors
will be studied on.
III.5.2.2. The Quality of Death Related Data
IMR and CMR levels are very important not only for the researchers and public
health specialists but for General Directorate of Mother and Child Health and Family
Planning of the Ministry of Health staff for policy purposes. Assessment of the
quality of the data used to estimate these rates directly from TDHSs survey is also
very important. Besides evaluating birth history data on the potential error points
mentioned above, the problems on the information about the age of the child at death
have to be studied. Birth history data is also crucial for the estimation of IMR and
CMR. Completeness and the accuracy of the data and its impacts on infant mortality
57
estimations are studied. On the other hand, in this study, the completeness and the
accuracy of the age at death data and its implications on mortality estimates are
studied. Same tests of data quality are going to be applied to all three surveys so that
the general quality of the data sets can be assessed and a comparison within these
surveys can be done.
The data quality of the birth history data -in general- is very important for
demographers to understand the accuracy and reliability of the results that they are
dealing with. As mentioned above, to follow the impact of the policies and develop
new ones, the data quality is vital. This study aims to fill a gap in the field of
demographic study in Turkey on data quality of TDHSs.
III.5.2.2.1. Date of Birth Data
The mortality estimations are based on the date of birth information. Hence, the
quality of the birth history data has direct effect on the estimations of the IMR and
CMR. The previous studies indicated that the quality of the data for the dead children
is worse than the living children. Mother either don’t remember or don’t want to
remember the exact information on the birth and death date of their children. The
DHS data set has a flag information for the birth date for the cases the imputation is
done. This indicates the quality difference between the living children and the dead
children birth data. The completeness of the information on the birth dates of living
children and the dead children is done for the type of place of residence, region,
mother’s age and education.
Moreover, the distribution of the death births according to the years prior to the
survey will show the extent of the displacement of the births. The displacement will
lead to a miscalculation of infant and child mortality estimations. In DHS reports
mortality rates are estimated for five year preceding the survey and if the
displacement is noticeable, the rates will be affected. The birth ratios for the fifth
calendar year preceding the survey by survival status of children is evaluated by
58
region, type of place of residence, age and education of the mother and the time
period of the interviewer at the field.
III.5.2.2.2. Age at Death Data
At the standard DHS questionnaire the age at death data is collected by a question
with two parts. The answer of the respondent is coded by the interviewer either in
days, months or years as units depending on the answer and in numbers regarding on
the units according to the recording rules. If the time mentioned is less than 1 month
it is coded as days, if it is 1 month to 2 years it is coded as months and if the time is 2
years or more the answer is coded as in years.
During the training it is stressed to the interviewers try their best to get the complete
information about the birth and death dates of the children. A special emphasis to the
correct and consistent death information occurred inside the 5 year preceding the
interview date was mentioned to be done by the interviewer. The standard DHS data
set has two variables keeping the date of death of children. First variable keeps the
information as they are recorded at the field and the second one is the converted time
of death information into months. For some cases, time of death answer can be
inconsistent with other answers and needed to be imputed. While imputing new
information is assigned both controlling other answers. For some cases the
interviewer might either not ask the question or forgot to record the answer, and
during data entry a date can be assigned for these cases with controlling the other
answers.
At this study, the completeness of time of death variable is evaluated in terms of data
quality. If the incompleteness of the information is high this will create a question for
the quality of the data and the situation in the field. The completeness of the time of
death is estimated in birth cohorts by socio-demographic characteristics of mother.
59
The accuracy of the data is also assessed in terms of heaping on 12th month at age of
death information. The index of heaping for month 12 is estimated by taking the age
at death information for months 10, 11, 13, and 14. The formula for Index of
Heaping is:
10 11 13 14
4 * 12
d d d d
IH d
+ + +
=
Where d= deaths on age x.
III.5.2.3. The Impact of Data Quality on Demographic Rates
III.5.2.3.1. Fertility Impact of Data Quality
The exclusion of the eligible women from the eligibility criterions will have a direct
effect on rates which are based on the information gathered from the women
questionnaire. Total fertility rate (TFR) is one of the important indicators of fertility
and can be calculated directly from the data collected from DHS. The interview date,
the birth date of woman and the birth date of the child information is used to estimate
TFR. TFR is a age period fertility rate for a synthetic cohort of women and measures
the average number of births a group of women would have by the time they reach
age 50 if they were to give birth at the current age-specific fertility rates and
expressed as the average number of births per woman (Rutstein and Rojas, 2003).
Standard DHS reports have been publishing the TFR either for 5 year or 3 years or
both. The quality of the dates has direct effect on the TFR and therefore is evaluated
at this study related with other variables.
At this study simulations which were aimed to show the effect of exclusion of
women are done. First group of simulations are made to estimate the effect of
60
exclusion of women on TFR. The estimated Boundary Effects and Sleeping Away
Exclusions were used for the simulations. The simulations are:
Total Fertility Rate Simulations
Simulations based on Boundary Effects
Simulation based on Lower Boundary Effect (FLB0)
Excluded women had an age-specific fertility of 0.0 (no births).
Simulation based on Upper boundary Effect - 1 (FUB0)
Excluded women had an age-specific fertility of 0.0 (no births).
Simulation based on Upper boundary Effect - 2 (FUB2)
Excluded women had twice the age-specific fertility as included.
Simulations based on Sleeping Away Exclusion
Simulations based on Sleeping Away Exclusion - 1 (FSA75)
Excluded woman had 75 percent of the age-specific fertility of included women.
Simulations based on Sleeping Away Exclusion - 2 (FSA125)
Excluded woman had 125 percent of the age-specific fertility of included women.
III.5.2.3.2. Mortality Impact of Data Quality
The impact of the exclusion of women at the household questionnaire caused by the
boundary effects and sleeping away is evaluated by using three simulations to
understand the impact of these exclusions on under-five mortality rate. In addition,
the heaping on month 12 at age at death data at birth history section of the ever
married women data is assessed for the impact of the heaping on infant and child
mortality rates.
61
Under-five Mortality Rate (5q0) Simulations
The impact of exclusion of eligible women to non eligible ages and the nexclusion
by sleeping away factoron under-five mortality rate is studied at this part of the
study.
Simulations based on Boundary Effects
Simulation based on Lower Boundary Effect (MLB150)
Excluded children have 150 percent the rate under-five mortality by age of mother as
included children.
Simulation based on Upper boundary Effect (MUB150)
Excluded children have 150 percent the rate under-five mortality by age of mother as
included children.
Simulation based on Sleeping Away Exclusion
Simulations based on Sleeping Away Exclusion (MSA150)
Excluded children have 150 percent the rate under-five mortality by age of mother as
included children.
The Effect of Heaping on 12th Month on Infant Mortality Rates
The age specific mortality estimates will be affected by the quality time of death
data. At that point, the importance of the heaping the death of children to 12 month,
comes with the calculation of Infant Mortality Rate (IMR) and Child Mortality Rate
(CMR) indicators, which counts in the deaths occurred until 1 year period and from 1
year to 5 year respectively. If there is a noticeable heaping for the 12 month at the
data, the quality of IMR and CMR will decrease. During data collection, a heaping to
1 month or 1 year for the time of death is expected. Interviewers are told to be
careful at such kind of heaping and were advised to probe to avoid these heaping. If
the respondent gives 1 month answer, the interviewer should probe and be sure that
62
the answer is exactly 1 month. Likewise, for the answers 1 year, the interviewer
should probe to get the exact month information. At this study, an evaluation of the
frequency of the 12 month answers for the time of death question in birth cohorts and
by socio demographic characteristics of mother is done. In addition, the effect of
heaping on the 12th month is evaluated on the infant mortality rates for all three
surveys at this study. The excess deaths calculated for the 12th month is redistributed.
The excess deaths are accepted as the difference between the number of deaths at 12
months and the average number at months 10, 11, 13 and 14 (Sullivan, et al., 1990).
The 25 % of the excess deaths are distributed to the months 0-11 from the month 12.
New IMR and CMR estimates are estimated with the new distribution of excess
deaths at month 12.
63
IV. THE ASSESSMENT OF DATA USED TO DETERMINE ELIGIBILITY
FOR THE INDIVIDUAL INTERVIEW
The data used to determine the eligibility for the individual interview is aimed to be
evaluated at this section of the study. Assessing the quality of the data used for the
selection for the individual interview is crucial for evaluating the overall quality of
the TDHS data. The selection process at the household data effects all means of rates
and ratios calculated. To study the effect of the data quality on fertility and mortality
rates, it is better to start with the household age and sex data.
The standard DHS household and individual questionnaires are applied in an order
that the interviewer starts the interview in a household by applying the household
questionnaire in which the household members are listed and various demographic
and social characteristics of the members are collected with the household
characteristics and assets in the household. The additional modules at the household
questionnaire used in TDHSs are never married and elderly modules to get
information for the never married 15-49 aged women and elderly people (aged 65+
and aged 60+ respectively at TDHS-1998 and TDHS-2003).
IV.I. The Assessment of the Data at Household Interview
IV.1.1. Household Interview Results
The household response rates are evaluated as an indicator for the data quality. The
lower the response rates the quality of the data can be criticized. The potential
eligible women at the households that cannot be reached or the questionnaire is not
filled will create a bias at the data. The sample weight calculated from the household
64
response rates accepts the interviewed households reflect the characteristics of the
not interviewed ones. However, the low level of completion rate will question the
quality of the data. Therefore, the household result rates are evaluated at this study as
an indicator of the data quality in terms of selection of the eligible women.
The detailed household results are evaluated at TDHS-1993, TDHS-1998 and TDHS-
2003 in general and by the region and type of place of residence for each survey.
Table IV.1.1., Table IV.1.2. and Table IV.1.3. present the household result codes and
response rates at TDHS-1993, TDHS-1998 and TDHS-2003 respectively. The
overall results indicate that the completed household rates are more than 80 percent
for all surveys. The highest rate is seen in TDHS-2003 with 83.0 percent followed by
TDHS-1993 (81.1 %) and TDHS-1998 (80.8). The highest completion rate is seen in
rural areas for all three surveys. The gap between the urban and rural is closing
around 2 percent between the legs of the TDHSs. The gap was 10 % at TDHS-1993
closed to 8.1 % at TDHS-1998 and at TDHS-2003. On the other hand, the difference
between the rural and urban was estimated as 5.8 %. Refusal has always been lower
in rural areas. The household results for all three surveys indicate that the refusals in
rural areas are very low as compared to the urban areas. Although the refusal rate
increased to 3.4 percent for Turkey, the refusal rate in rural areas for this survey
dropped to 0.4 percent when compared with the TDHS-1998 (0.6 %).
When the completion rates are evaluated by regions, it is seen that, although one of
the lowest completion rates are seen at East region, the lasts TDHS indicated that 9
out of 10 of the household questionnaires were completed. The refusal rate on the
other hand seems increased in West region at TDHS-2003 (6.3 %). At this survey the
lowest refusal is seen in East region. It is clearly seen that from one survey to another
the refusal rates are changing and it is hard to mention a trend for the refusal rates
between the surveys.
The “dwelling destroyed” and “dwelling not found” result codes have direct relation
with the listing procedure which is done seriously in TDHSs. The blocks selected
from the address lists of TURKSTAT (former name was State Institute of Statistics
65
(SIS)) were visited by listing personnel of HUIPS. The main aim was to update the
dwellings at the block selected. The systematic selection is done after the complete
listing of the block is done. The dwellings in which there is no households are living
is not included in the selection process. This listing procedure aims to increase the
response rates and decrease the rates for “dwelling destroyed” and “dwelling not
found” as result codes. With a successful listing operation for all three surveys the
total percent of these two codes are always below 2 %. The lowest rates for them are
seen in TDHS-2003 with 0.3 %.
The response rates are estimated from the raw data for all three surveys. Mainly
because of the increasing refusal rate, the response rates are decreasing among three
surveys. On the other hand, there is a clear difference among urban and rural
response rates. For all three surveys the response rates for rural areas are above 97
percent (99.4 % at TDHS-1993, 97.0 % at TDHS-1998 and 98.0 % at TDHS-2003).
However, the response rates for urban households decreased from 95.5% at TDHS-
1993 to 91.4 % at TDHS-2008. The acceptance for the interview at the urban areas
seems decreasing for the interviewers.
The regional estimation for response rates reveals the clear decline at the West region
from TDHS-1998 to TDHS-2003 (941 % to 88.6 %). It is important to stress the fact
that the only response rate below 90% has seen for all three surveys is at the West
region at TDHS-2003. The gap between the region having highest response rate and
the lowest increased at this survey to around 8 %. This gap was around 4 % for the
previous surveys. On the other hand, the highest response rates are estimated at
South region for the first two DHSs. While at TDHS-2003, the highest response rates
are seen at East region (96.5%), the highest rate of all three surveys is seen at South
region in TDHS-1993 (98.7%).
66
Table IV.1.1. Household Response Rates and Percent Distribution of Household Result Codes by Region and Type of Place of
Residence, TDHS-1993
Completed
HH present,
no resp.
HH
absent Postponed Refused
Dwelling
vacant
Dwelling
destroyed
Dwelling
not found Other Total
Number of
Households
Household
Response Rate
Region
West 79.2 0.2 9.8 0.1 3.7 6.2 0.3 0.3 0.2 100.0 3,374 94.9
South 84.6 0.3 7.9 0.0 0.5 5.7 0.3 0.3 0.2 100.0 2,045 98.7
Central 85.1 0.4 8.6 0.0 1.3 4.1 0.3 0.1 0.1 100.0 2,269 97.9
North 76.3 0.5 16.0 0.0 0.5 6.1 0.3 0.2 0.1 100.0 1,554 98.5
East 79.0 0.2 10.7 0.1 1.9 5.5 0.6 1.7 0.4 100.0 1,389 95.3
Type of place of residence
Urban 77.7 0.3 11.8 0.0 2.8 6.3 0.4 0.6 0.1 100.0 7,065 95.5
Rural 87.7 0.3 7.0 0.0 0.1 4.1 0.3 0.1 0.4 100.0 3,566 99.4
Total 81.1 0.3 10.2 0.0 1.9 5.6 0.3 0.4 0.2 100.0 10,631 96.9
67
Table IV.1.2. Household Response Rates and Percent Distribution of Household Result Codes by Region and Type of Place of
Residence, TDHS-1998
Completed
HH present,
no resp.
HH
absent Postponed Refused
Dwelling
vacant
Dwelling
destroyed
Dwelling
not found Other Total
Number of
Households
Household
Response Rate
Region
West 83.9 3.1 6.0 0.5 1.4 4.8 0.0 0.3 0.0 100.0 2,827 94.1
South 83.3 2.5 5.6 0.2 1.5 6.7 0.1 0.2 0.0 100.0 1,815 95.0
Central 80.3 4.3 7.2 0.3 2.5 5.0 0.1 0.2 0.0 100.0 2,104 91.7
North 74.8 2.7 9.6 0.1 1.2 9.9 0.1 1.4 0.1 100.0 1,479 93.3
East 79.1 0.9 8.5 0.1 1.1 7.0 0.4 2.2 0.7 100.0 1,745 94.8
Type of place of residence
Urban 78.4 3.3 7.8 0.3 2.0 6.9 0.1 1.0 0.2 100.0 6,989 92.2
Rural 86.5 1.7 5.6 0.2 0.6 5.1 0.1 0.2 0.0 100.0 2,981 97.0
Total 80.8 2.8 7.2 0.3 1.6 6.3 0.1 0.8 0.2 100.0 9,970 93.6
68
Table IV.1.3. Household Response Rates and Percent Distribution of Household Result Codes by Region and Type of Place of
Residence, TDHS-2003
Completed
HH present,
no resp.
HH
absent Postponed Refused
Dwelling
vacant
Dwelling
destroyed
Dwelling
not found Other Total
Number of
Households
Household
Response Rate
Region
West 79.0 3.4 4.7 0.1 6.3 5.5 0.0 0.4 0.6 100,0 4,267 88.6
South 85.1 3.1 4.2 0.0 2.4 4.6 0.0 0.6 0.1 100,0 1,797 93.3
Central 83.0 2.1 8.0 0.1 2.0 4.6 0.0 0.0 0.2 100,0 2,433 95.2
North 80.2 1.8 9.9 0.0 1.8 6.0 0.0 0.3 0.1 100,0 1,587 95.4
East 89.2 1.3 3.6 0.0 1.7 3.6 0.0 0.2 0.2 100,0 2,965 96.5
Type of place of
residence
Urban 81.6 2.9 5.0 0.1 4.4 5.4 0.0 0.3 0.3 100,0 9,754 91.4
Rural 87.4 1.2 7.4 0.0 0.4 3.1 0.0 0.2 0.3 100,0 3,295 98.0
Total 83.0 2.5 5.6 0.0 3.4 4.8 0.0 0.3 0.3 100,0 13,049 93.0
69
IV.1.2. The Quality of Age Reporting in Household Questionnaire
The two indicators for the eligibility of the individual interview are sex and age. At
the training period and during the field survey, interviewers were told to record the
age and sex of the household members as correct as possible. At the household
questionnaire with one question “How old is …….?” the age information for all
household members is collected from the answers of a member of the household as
proxy informant. The single age distribution of the de facto male, female and total
household members at TDHS-1993 are presented at Table IV.1.2.1. and Figure
IV.1.2.1.. It is clearly seen that there is a digit preference for 0 and 5. Digit
preference seems clearer among females as compared to males. As the results are
evaluated focusing on the eligibility criterion ages for females (ages 14, 15 and 49,
50), it is seen that there is no intensification for age15 but a shift or a heaping can
come to mind for age 50. For the first 14 ages the percent of male household
members are higher than the female. This can be a reflection of the usual result of
sex ratio at birth (usually 105 male births for 1000 female births) or the traditions to
mention male children and ignore the female ones. This trend can be also seen at the
other two survey age distributions.
The de facto age distribution of household members at TDHS-1998 is presented at
Table IV.1.2.2. and Figure IV.1.2.2.. The percent difference between male and
female household members to age 15 is clear for also this survey. On the other hand,
the heaping for ages ending with 0 or 5 is also evident. The sharp fluctuations at the
figure show the extent of the digit preference. Similar to TDHS-1993, TDHS-1998
and TDHS-2003 results shows that digit preference is more common among female
household members. On the other hand, Table IV.1.2.3 indicates the de facto
household population for TFHS-2003. Similar to previous surveys, a clear digit
preference for ages ending with 0 and 5 is also seen at TDHS-2003.
The household member’s age distribution shows some fluctuations which cannot be
only explained by digit preference or heaping problem. For all three surveys there is
a sharp decrease for ages around 20 for male members. The sample design of the
70
TDHS does not include the institutional population like students at the dormitories,
prisoners or soldiers. As the military service is obligatory for males after age 18, the
noticeable fluctuation for males around age 20 can be a reflection of the soldier
population at these households which are not mentioned as a member.
71
Table IV.1.2.1. De Facto Age Distribution of TDHS-1993
Male Female Total Male Female Total
Age n % n % n % Age n % n % n %
0 396 2.12 373 1.91 769 2.01 50 226 1.21 198 1.01 424 1.11
1 328 1.75 315 1.61 643 1.68 51 105 0.56 169 0.86 274 0.72
2 329 1.76 323 1.65 652 1.70 52 118 0.63 209 1.07 327 0.85
3 365 1.95 325 1.66 690 1.80 53 150 0.80 191 0.98 341 0.89
4 374 2.00 355 1.81 729 1.90 54 104 0.56 130 0.66 234 0.61
5 362 1.93 355 1.81 717 1.87 55 248 1.33 285 1.46 533 1.39
6 385 2.06 415 2.12 800 2.09 56 127 0.68 126 0.64 253 0.66
7 468 2.50 444 2.27 912 2.38 57 100 0.53 126 0.64 226 0.59
8 481 2.57 468 2.39 949 2.48 58 126 0.67 116 0.59 242 0.63
9 467 2.50 392 2.00 859 2.24 59 87 0.46 76 0.39 163 0.43
10 517 2.76 499 2.55 1016 2.65 60 269 1.44 327 1.67 596 1.56
11 450 2.41 419 2.14 869 2.27 61 95 0.51 64 0.33 159 0.42
12 521 2.78 511 2.61 1032 2.70 62 106 0.57 99 0.51 205 0.54
13 546 2.92 497 2.54 1043 2.72 63 116 0.62 108 0.55 224 0.59
14 446 2.38 472 2.41 918 2.40 64 72 0.38 78 0.40 150 0.39
15 444 2.37 471 2.41 915 2.39 65 222 1.19 239 1.22 461 1.20
16 446 2.38 498 2.54 944 2.47 66 82 0.44 116 0.59 198 0.52
17 439 2.35 521 2.66 960 2.51 67 85 0.45 97 0.50 182 0.48
18 428 2.29 492 2.51 920 2.40 68 65 0.35 57 0.29 122 0.32
19 344 1.84 382 1.95 726 1.90 69 41 0.22 27 0.14 68 0.18
20 298 1.59 453 2.31 751 1.96 70 144 0.77 145 0.74 289 0.75
21 221 1.18 352 1.80 573 1.50 71 21 0.11 23 0.12 44 0.11
22 311 1.66 389 1.99 700 1.83 72 32 0.17 37 0.19 69 0.18
23 358 1.91 360 1.84 718 1.88 73 39 0.21 24 0.12 63 0.16
24 310 1.66 317 1.62 627 1.64 74 14 0.07 19 0.10 33 0.09
25 360 1.92 355 1.81 715 1.87 75 64 0.34 66 0.34 130 0.34
26 261 1.39 284 1.45 545 1.42 76 18 0.10 21 0.11 39 0.10
27 297 1.59 315 1.61 612 1.60 77 16 0.09 12 0.06 28 0.07
28 298 1.59 298 1.52 596 1.56 78 16 0.09 16 0.08 32 0.08
29 228 1.22 221 1.13 449 1.17 79 17 0.09 6 0.03 23 0.06
30 355 1.90 382 1.95 737 1.93 80 54 0.29 71 0.36 125 0.33
31 203 1.08 219 1.12 422 1.10 81 9 0.05 3 0.02 12 0.03
32 221 1.18 246 1.26 467 1.22 82 15 0.08 17 0.09 32 0.08
33 250 1.34 332 1.70 582 1.52 83 6 0.03 10 0.05 16 0.04
34 203 1.08 218 1.11 421 1.10 84 5 0.03 9 0.05 14 0.04
35 305 1.63 292 1.49 597 1.56 85 24 0.13 29 0.15 53 0.14
36 212 1.13 199 1.02 411 1.07 86 4 0.02 10 0.05 14 0.04
37 250 1.34 212 1.08 462 1.21 87 11 0.06 10 0.05 21 0.05
38 258 1.38 271 1.38 529 1.38 88 1 0.01 7 0.04 8 0.02
39 187 1.00 183 0.94 370 0.97 89 2 0.01 3 0.02 5 0.01
40 285 1.52 297 1.52 582 1.52 90 9 0.05 23 0.12 32 0.08
41 134 0.72 149 0.76 283 0.74 91 2 0.01 2 0.01 4 0.01
42 183 0.98 188 0.96 371 0.97 92 1 0.01 2 0.01 3 0.01
43 221 1.18 224 1.14 445 1.16 93 4 0.02 1 0.01 5 0.01
44 130 0.69 135 0.69 265 0.69 94 1 0.01 0 0.00 1 0.00
45 227 1.21 217 1.11 444 1.16 95 12 0.06 16 0.08 28 0.07
46 125 0.67 139 0.71 264 0.69 DK 3 0.02 4 0.02 7 0.02
47 130 0.69 134 0.68 264 0.69 Missing 5 0.03 1 0.01 6 0.02
48 158 0.84 157 0.80 315 0.82 Total 18710 100.00 19571 100.00 38281 100.00
49 102 0.55 81 0.41 183 0.48
72
Table IV.1.2.2. De Facto Age Distribution of TDHS-1998
Male Female Total Male Female Total
Age n % n % n % Age n % n % n %
0 387 2.35 367 2.12 754 2.23 50 172 1.04 175 1.01 347 1.03
1 377 2.29 335 1.93 712 2.11 51 106 0.64 149 0.86 255 0.75
2 344 2.09 304 1.75 648 1.92 52 139 0.84 119 0.69 258 0.76
3 333 2.02 310 1.79 643 1.90 53 106 0.64 127 0.73 233 0.69
4 349 2.12 339 1.95 688 2.04 54 97 0.59 115 0.66 212 0.63
5 336 2.04 316 1.82 652 1.93 55 148 0.90 195 1.12 343 1.01
6 409 2.48 388 2.24 797 2.36 56 93 0.56 118 0.68 211 0.62
7 346 2.10 350 2.02 696 2.06 57 103 0.63 92 0.53 195 0.58
8 403 2.45 369 2.13 772 2.28 58 114 0.69 119 0.69 233 0.69
9 307 1.87 335 1.93 642 1.90 59 78 0.47 95 0.55 173 0.51
10 389 2.36 323 1.86 712 2.11 60 197 1.20 193 1.11 390 1.15
11 361 2.19 340 1.96 701 2.07 61 64 0.39 48 0.28 112 0.33
12 361 2.19 378 2.18 739 2.19 62 96 0.58 74 0.43 170 0.50
13 386 2.34 379 2.19 765 2.26 63 72 0.44 86 0.50 158 0.47
14 322 1.96 403 2.32 725 2.14 64 75 0.46 70 0.40 145 0.43
15 329 2.00 330 1.90 659 1.95 65 165 1.00 163 0.94 328 0.97
16 347 2.11 421 2.43 768 2.27 66 79 0.48 94 0.54 173 0.51
17 370 2.25 378 2.18 748 2.21 67 65 0.39 79 0.46 144 0.43
18 404 2.45 460 2.65 864 2.56 68 63 0.38 81 0.47 144 0.43
19 307 1.87 318 1.83 625 1.85 69 48 0.29 42 0.24 90 0.27
20 240 1.46 380 2.19 620 1.83 70 111 0.67 143 0.82 254 0.75
21 203 1.23 320 1.85 523 1.55 71 52 0.32 30 0.17 82 0.24
22 290 1.76 330 1.90 620 1.83 72 43 0.26 54 0.31 97 0.29
23 291 1.77 324 1.87 615 1.82 73 35 0.21 40 0.23 75 0.22
24 302 1.83 347 2.00 649 1.92 74 32 0.19 28 0.16 60 0.18
25 325 1.97 328 1.89 653 1.93 75 58 0.35 56 0.32 114 0.34
26 304 1.85 338 1.95 642 1.90 76 28 0.17 27 0.16 55 0.16
27 266 1.62 276 1.59 542 1.60 77 17 0.10 13 0.07 30 0.09
28 254 1.54 285 1.64 539 1.59 78 29 0.18 25 0.14 54 0.16
29 206 1.25 256 1.48 462 1.37 79 9 0.05 16 0.09 25 0.07
30 266 1.62 291 1.68 557 1.65 80 23 0.14 43 0.25 66 0.20
31 193 1.17 180 1.04 373 1.10 81 2 0.01 8 0.05 10 0.03
32 225 1.37 251 1.45 476 1.41 82 5 0.03 10 0.06 15 0.04
33 255 1.55 277 1.60 532 1.57 83 6 0.04 11 0.06 17 0.05
34 236 1.43 256 1.48 492 1.46 84 8 0.05 10 0.06 18 0.05
35 240 1.46 237 1.37 477 1.41 85 16 0.10 19 0.11 35 0.10
36 199 1.21 229 1.32 428 1.27 86 7 0.04 11 0.06 18 0.05
37 215 1.31 236 1.36 451 1.33 87 5 0.03 10 0.06 15 0.04
38 249 1.51 262 1.51 511 1.51 88 2 0.01 10 0.06 12 0.04
39 160 0.97 189 1.09 349 1.03 89 1 0.01 0 0.00 1 0.00
40 255 1.55 239 1.38 494 1.46 90 5 0.03 16 0.09 21 0.06
41 156 0.95 162 0.93 318 0.94 91 2 0.01 0 0.00 2 0.01
42 218 1.32 191 1.10 409 1.21 92 0 0.00 5 0.03 5 0.01
43 182 1.11 209 1.21 391 1.16 93 1 0.01 4 0.02 5 0.01
44 168 1.02 174 1.00 342 1.01 94 1 0.01 3 0.02 4 0.01
45 193 1.17 206 1.19 399 1.18 95 2 0.01 10 0.06 12 0.04
46 160 0.97 160 0.92 320 0.95 DK 14 0.09 8 0.05 22 0.07
47 124 0.75 122 0.70 246 0.73 Total 16461 100.00 17341 100.00 33802 100.00
48 197 1.20 197 1.14 394 1.17
49 128 0.78 102 0.59 230 0.68
73
Table IV.1.2.3. De Facto Age Distribution of TDHS-2003
Male Female Total Male Female Total
Age n % n % n % Age n % n % n %
0 399 1.91 370 1.68 769 1.79 50 235 1.13 198 0.90 433 1.01
1 354 1.70 348 1.58 702 1.64 51 152 0.73 193 0.88 345 0.81
2 447 2.14 404 1.84 851 1.99 52 189 0.91 240 1.09 429 1.00
3 479 2.30 414 1.88 893 2.08 53 191 0.92 243 1.10 434 1.01
4 403 1.93 404 1.84 807 1.88 54 165 0.79 196 0.89 361 0.84
5 409 1.96 408 1.85 817 1.91 55 212 1.02 220 1.00 432 1.01
6 421 2.02 408 1.85 829 1.93 56 145 0.70 125 0.57 270 0.63
7 438 2.10 432 1.96 870 2.03 57 125 0.60 139 0.63 264 0.62
8 447 2.14 439 1.99 886 2.07 58 161 0.77 169 0.77 330 0.77
9 400 1.92 412 1.87 812 1.89 59 142 0.68 91 0.41 233 0.54
10 412 1.98 433 1.97 845 1.97 60 132 0.63 185 0.84 317 0.74
11 427 2.05 439 1.99 866 2.02 61 110 0.53 83 0.38 193 0.45
12 416 2.00 376 1.71 792 1.85 62 97 0.47 107 0.49 204 0.48
13 450 2.16 453 2.06 903 2.11 63 120 0.58 133 0.60 253 0.59
14 414 1.99 424 1.93 838 1.96 64 97 0.47 128 0.58 225 0.53
15 386 1.85 387 1.76 773 1.80 65 173 0.83 220 1.00 393 0.92
16 450 2.16 445 2.02 895 2.09 66 81 0.39 97 0.44 178 0.42
17 456 2.19 468 2.13 924 2.16 67 81 0.39 120 0.55 201 0.47
18 414 1.99 435 1.98 849 1.98 68 75 0.36 91 0.41 166 0.39
19 372 1.78 369 1.68 741 1.73 69 67 0.32 77 0.35 144 0.34
20 246 1.18 442 2.01 688 1.61 70 172 0.83 177 0.80 349 0.81
21 303 1.45 364 1.65 667 1.56 71 60 0.29 61 0.28 121 0.28
22 381 1.83 484 2.20 865 2.02 72 63 0.30 74 0.34 137 0.32
23 397 1.90 435 1.98 832 1.94 73 78 0.37 87 0.40 165 0.39
24 369 1.77 411 1.87 780 1.82 74 50 0.24 58 0.26 108 0.25
25 362 1.74 409 1.86 771 1.80 75 90 0.43 115 0.52 205 0.48
26 327 1.57 412 1.87 739 1.72 76 57 0.27 67 0.30 124 0.29
27 304 1.46 343 1.56 647 1.51 77 55 0.26 44 0.20 99 0.23
28 324 1.55 374 1.70 698 1.63 78 46 0.22 38 0.17 84 0.20
29 319 1.53 337 1.53 656 1.53 79 37 0.18 22 0.10 59 0.14
30 375 1.80 414 1.88 789 1.84 80 48 0.23 70 0.32 118 0.28
31 286 1.37 305 1.39 591 1.38 81 19 0.09 13 0.06 32 0.07
32 329 1.58 311 1.41 640 1.49 82 20 0.10 20 0.09 40 0.09
33 299 1.43 334 1.52 633 1.48 83 9 0.04 25 0.11 34 0.08
34 284 1.36 283 1.29 567 1.32 84 15 0.07 16 0.07 31 0.07
35 307 1.47 325 1.48 632 1.47 85 19 0.09 16 0.07 35 0.08
36 227 1.09 224 1.02 451 1.05 86 8 0.04 14 0.06 22 0.05
37 249 1.19 298 1.35 547 1.28 87 4 0.02 6 0.03 10 0.02
38 324 1.55 327 1.49 651 1.52 88 4 0.02 1 0.00 5 0.01
39 271 1.30 334 1.52 605 1.41 89 1 0.00 6 0.03 7 0.02
40 335 1.61 306 1.39 641 1.50 90 18 0.09 15 0.07 33 0.08
41 203 0.97 262 1.19 465 1.09 91 2 0.01 3 0.01 5 0.01
42 260 1.25 274 1.25 534 1.25 92 0 0.00 3 0.01 3 0.01
43 284 1.36 316 1.44 600 1.40 93 4 0.02 5 0.02 9 0.02
44 256 1.23 250 1.14 506 1.18 94 1 0.00 1 0.00 2 0.00
45 280 1.34 287 1.30 567 1.32 95 13 0.06 14 0.06 27 0.06
46 195 0.94 215 0.98 410 0.96 DK 17 0.08 4 0.02 21 0.05
47 253 1.21 225 1.02 478 1.12 Total 20843 100.00 22007 100.00 42850 100.00
48 235 1.13 242 1.10 477 1.11
49 205 0.98 166 0.75 371 0.87
74
Figure IV.1.2.1. Age Distribution of De Facto Household Population,
TDHS-1993
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Male Female TDHS-1993
Figure IV.1.2.2. Age Distribution of De Facto Household Population,
TDHS-1998
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Male Female
TDHS-1998
75
Figure IV.1.2.3. Age Distribution of De Facto Household Population,
TDHS-2003
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Male Female TDHS-2003
At this study, the Myers, Whipple, Bachi and UN Age-Sex Ratio Indexes were
estimated in terms of assessing the data quality of the household members at TDHS-
1993, TDHS-1998 and TDHS-2003. The Myers digit preference at these three
surveys can easily be identified with figures from Figure IV.1.2.4. to IV.1.2.6.. For
all three surveys, the obvious digit preference problem for ages ending with “0” and
“5” is seen. For male population the preference for these ages are lower than females.
However, the results indicate that the problem of digit preference is decreasing. The
Myers digit preference for “0” was 3.4 for males and 4.1 for females at TDHS-1993.
However, it dropped down to 1.5 and 2.0 at TDHS-1998 and 0.8 and 1.5 at TDHS-
2003 respectively for males and females. On the other hand, Myers estimation for the
preference of digit “5” indicates that at TDHS-1993 and TDHS-1998 the preference
is seen more for males than females. At TDHS-2003 the preference of digit “5” is
more or less same for males and females. Likewise, the preference of “5” decreased
among three surveys.
76
In addition to “0” and “5”, the digits “3” and “8” seems attractive for the
respondents. The TDHSs have applied at the years ending with 3 and 8. The
respondents may think the birth year of the household member and they may round
up the year of birth and calculate the age with using this rounded year information.
Except TDHS-1998, the digit preference for “3”and “8” is remarkable. At TDHS-
1998, although there is no clear preference for digit “3”, a vivid digit preference for
“8” is seen. Even this preference is close to the preference of “0” and “5” for this
survey.
On the other hand, the preference of certain digits causes problem of “nonpreferable”
digits like “1”, “4”, “6” and “9”. The figures indicate that, for these digits
the preference estimation resulted in minus values. It is clear that, as the neighboring
digits are at the center of interest; these digits are less preferred by the respondents.
However, the overall preference and non- preference seems decreasing for males and
females and total population. With this survey the digit preference for Bachi Index is
also estimated and presented at Annex VIII.with Figure VIII.2., Figure VIII.2.2. and
Figure VIII.2.3.. As the results of the index shows little dissimilarity from the Myers
index estimations, no additional comments were done on the results. Interested
readers may find the results at this section of the study.
77
Figure IV.1.2.4. Myers Preference by Digit, TDHS-1993
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
0 1 2 3 4 5 6 7 8 9
TURKEY: TDHS-1993 Total
Male Female
Myers Preference by Digit
Figure IV.1.2.5. Myers Preference by Digit, TDHS-1998
-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
0 1 2 3 4 5 6 7 8 9
TURKEY: TDHS-1998 Total
Male Female
Myers Preference by Digit
78
Figure IV.1.2.6. Myers Preference by Digit, TDHS-2003
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
0 1 2 3 4 5 6 7 8 9
TURKEY: TDHS-2003 Total
Male Female
Myers Preference by Digit
Table IV.1.2.4 indicates the household age and sex ratios and United Nations Index
at TDHS-1993, 1998 and TDHS-2003 and Myers, Bachi and Whipple Indices for
female household population for these surveys. The table presents the estimations by
region and type of place of residence and total. As the main focus of this section is to
assess the data quality of the data used for the eligibility of women questionnaire,
although the index results are calculated for total population, the Table VIII.2.1. at
Annex VIII.2. indicates the results for both sexes and they are not going to be
discussed here.
The Myers index results for females show that the age heaping problem is decreasing
at TDHSs. The Index results at TDHS-1993 and TDHS-1998 indicates that the digit
preference problem for these surveys among female population is at medium level.
On the other hand, as the index score decreased to an 8.2 at TDHS-2003; the heaping
problem can be regarded as low and acceptable for this survey. There was clear
regional difference in terms of the level of digit preference problem. At TDHS-1993
the Myers index is estimated at 26.2 –which is the only result below medium level.
79
On the contrary, the results for other regions are at medium level for the same
survey. With TDHS-1998, the age heaping problem in West, Central and North
regions decreased to acceptable levels. In addition, the problem in East region (17.5)
drops to medium level for this survey. With TDHS-2003, for all regions including
East, the Myers index results are below 10. The lowest results were seen in different
regions for different surveys. While the lowest result is seen in West region (11.7) at
TDHS-1993, at TDHS-1998 the lowest result comes from the data of North region
(7.4). At TDHS-2003 the lowest Myers index result is calculated for South region
with a score of 6.8.
There is a gradual decrease in Myers Index results for both urban and rural female
population of Turkey. The results indicate that, at TDHS-1993 for both urban and
rural areas the age heaping problem is at medium level for females. With the TDHS-
1998 it is seen that for urban female population the size of the problem is calculated
low and acceptable (8.6)., whereas for the same survey for rural female population
the digit preference is still at medium level (16.3). With TDHS-2003, the digit
preference problem is estimated acceptable for both residential areas (7.2 and 9.9 for
urban and rural female populations respectively).
The Bachi index is another way of estimation for evaluating the age distribution
problem of populations. The results of Bachi Index for female population at TDHS-
1993, TDHS-1998 and TDHS-2003 is also shown in Table IV.1.2.4.. Bachi index
gives us results for ages 23 to 77 and the index scores are assessed same as Myers
index scores. As the age evaluated at this index is limited with the adult ages, the
scores are lower than Myers index scores. The respondents with whom the household
list is filled with possibly have better information about the age of the females aged
23 to 77 as compared with the all females. The score for Turkey decreased from
medium level with 11.0 at TDHS-1993 to low level 6.5 at TDHS-1998 and 5.2 at
TDHS-2003. At the urban areas Bachi index score have always lower than the rural
areas. For the last two surveys the score is even below 5 among females aged 23 to
77 living in urban areas. In rural areas, on the other hand, at TDHS-1993 and TDHS-
1998 the Bachi index is calculated as medium level and for the last TDHS, the score
80
dropped to accurate level (6.2). The distinction between East region and rest can be
easily seen for the Bachi index results at TDHS-1193. Although the highest score for
other regions is 11.9 at North, digit preference for East is estimated at 20.5 (high).
However, the problem of age heaping first decreased to medium level at TDHS-1998
and to low at TDHS-2003 at East region.
The results of the Whipple Index calculations are also presented at Table IV.1.2.4 for
Regions, Type of Place of Residence and Turkey. The overall results point to a
decrease at the level of age reporting error for Turkey among three surveys. At
TDHS-1993 the score is evaluated as “Rough”, at TDHS-1998 and at TDHS-2003 as
“Approximate”. If the trend continues, for the next survey the Whipple Index score
will be calculated as “Fairly Accurate”. The regional differences seen at other Index
scores can also be seen for Whipple Index score. The poorest (highest) and the best
(lowest) score among all three surveys is calculated for East Region at TDHS-1993
with 193 (Very rough) and at TDHS-2003 for North region with 99 (Highly accurate)
respectively. Like other index results, the quality of the urban female age data is in
good condition as compared to rural. For the last two surveys the Whipple index
score indicates fairly accurate age data for females.
The UN Age-Sex Index is also applied to all three TDHS data. The results indicate
that the overall joint score for Turkey for last two surveys is inaccurate whereas at
TDHS-1993 the score is highly inaccurate. The distribution of the males for
especially early ages and the heaping problem for certain ages creates such kind of
results. The males at military ages will also affect the score. For all regions and
surveys the UN index is calculated above 40 which show a highly inaccurate age and
sex distribution. One of the interesting findings is the data quality of the population
living in rural areas. At TDHS-1998 the joint score was very close to inaccurate level
(40.3) and at TDHS-2003 the score is estimated as inaccurate (38.4).
For the eligible age groups (15-49) and the previous age group of the first eligible
age group (10-14) and latter age group (50-54) is evaluated in terms of age and sex
ratio. When these age groups are compared regarding the age ratios of females, it is
81
seen that except last survey, the latter age group has more number of females as
compared to the former. However, at TDHS-2003, the 10-14 age group has more
females than the first eligible age group. Except West region at TDHS-2003, for
nearly all regions and surveys the age ratio for 10-14 age group is lower than 15-19
age group. For West region at TDHS-2003, there is a clear distinction between the
age ratios for these two age groups. This finds its reflection on the urban age ratio
results in TDHS-2003. While for the rest surveys and for urban and rural areas the
age ratio of the first eligible age group is higher than the previous age group.
When age ratios are assessed for the last eligible age group and the next, it is seen
that females are less mentioned at 45-49 age group when compared with the 50 -54.
Although there is no clear trend for the first eligible age group and the previous one,
for all three surveys, obvious trend can be seen for the last eligible age group and the
next. The data at TDHS-1998 is somehow shows different trend for age ratios for
these age groups. However, TDHS-1993 and TDHS-2003 results confirm the
imbalance at the age groups 45-49 and 50-54. This creates an exaggerated female
population for age group 50-54 and will result in problems in estimation of fertility
rates.
Table IV.1.2.4 also illustrates the sex ratios for the age groups 10-14, 15-19, 45-49
and 50-54 which give the information on the data quality in terms of carrying the
eligible women to not eligible age groups. It is seen that there is no clear problem of
underestimating the female population at first eligible age group for all TDHSs. The
female population at age group 15-19 is higher than male population. The only
situation where male population is higher than female is in TDHS-1993 for the age
group 10-14. At this survey, for this age group, around 103 males were mentioned
for 100 females.
The residential differentiation can be identified for nearly all surveys. For the first
two TDHSs, the sex ratio for age group 10-14 is very close to each other and above
104. Whereas, at TDHS-2003 the sex ratio for age group 10-14 is 97 and for the first
eligible age group it is estimated 107.6. This brings the possibility of carrying the
82
eligible females to not eligible ages into mind. For rural areas this kind of problem is
not seen, the only situation is for the TDHS-1998 where for both age groups the sex
ratios are 93.6 and 80.6 for age groups 10-14 and 15-19 respectively. When regional
sex ratio evaluation done for all three surveys, it is except the West region, there is
clear problem of carrying females to 10-14 age groups. At west for all surveys there
are more males at age group 15-19 as compared to age 10-14.
The sex ratio at the last eligible age group (45-49) and the next age group (50-54) is
also assessed at Table IV.1.2.4.. The results indicate that there are more females at
age group 50-54 when compare to 45-49. There were 78.4 males for 100 females at
50-59 ages at TDHS-1993, 90.5 at TDHS-1998 and 87 at the last TDHS. On the
other hand sex ratios for three TDHSs are calculated as 101.9, 101.9 and 102.9
respectively for the last eligible age group. Although there is a balance between the
sexes are seen for age group 45-49, women excess is seen for the age group 50-54.
This problem is more vivid in urban areas. While the sex ratio for the last eligible age
group is around 105 for all TDHSs, the sex ratio drops to 80.0, 98.5 and 87.5 for the
surveys in that order. In the rural areas, for both the last eligible age group and the
next age group women are mentioned more than males. However the excess of
female population at rural areas for the age group 50-54 is obvious. TDHS-1993
results indicate that for age group 50-54, for 100 females only 76.6 males are
mentioned. Although this increased to first 80.0 and 85.3 respectively for the last two
surveys, sex ratios are still questionable for rural areas.
The estimations for regions regarding the sex ratio for age groups 45-49 and 50-54
indicates that especially for East and Central region systematic repelling of women to
50-54 age group will be a problem for these regions. While 123.4 males were
mentioned at age group 45-49 at TDHS-1993 at East Region; for the same survey for
50-54 age group only 66.7 men were mentioned. Although the imbalance at the age
group 50-54 seems decreased with the surveys TDHS-1998 and TDHS-2003, the
problem at the last eligible age group continues. The imbalance of sexes for the 45-
49 and 50-54 age groups increased at Central region with the surveys. While there
were 102.3 males and 84.3 males mentioned for 100 females at age groups 45-49 and
83
50-54 respectively at TDHS-1993 the gap increased to 114.5 males and 79.8 males
per 100 females at TDHS-2003. It will be true to mention an overall excess of
females at 50-54 age group with respect to 45-49 nearly for all regions.
The indices, used for the assessment of the age data for females are assessed for each
TDHS data sets by the selected demographic characteristics of the respondent whom
the household interview is completed. The results are presented at Table IV.1.2.6. To
select the best respondent for the household interview is crucial for the data quality
of the survey. The results indicate that the best information is taken from either
household head or her/his spouse. For all three surveys the answers less affected by
age heaping and digit preference is brought by these members of the household. The
high index results are estimated for the household members who are other than
household head, wife/husband or son/daughter. The age of the respondent is
evaluated in terms of indices estimating the digit preference and age heaping. Results
show that when the respondent is above age 55, the quality of the data is very low.
The best answers were taken by the 35-54 aged respondents at TDHS-1993 and
TDHS-2003. On the other hand, at TDHS-1998, the best quality age information is
taken from the interviews done with 15-34 aged members of the household.
The female respondents seem giving more dirty age data in terms of age heaping and
digit preference as compared to males at the first and the last TDHSs. In TDHS-1998
females gave better quality age information as compared to males at the same survey.
Education seems the most important factor affecting the data quality of age at the
household list. Except TDHS 1993, the index results decrease with the increase at
education levels. The index results for the group of respondent have no education or
did not complete the primary education is two times higher than the index estimated
from the answers of respondents completed the primary education.
84
Table IV.1.2.4. Household Age and Sex Ratios and Myers, Bachi, Whipple and United Nations Indices for Household Data by
Region and Type of Place of Residence, TDHS 1993, 1998 and 2003
Myers Index
(Females)
Bachi Index
(Females)
Whipple Index
(Females)
UN
Index
Age Ratios (Females) Sex Ratios
10-14 15-19 45-49 50-54 10-14 15-19 45-49 50-54
Region
West
1993 11.7 7.9 123 54.6 114.0 103.8 79.7 123.7 97.8 104.2 99.3 78.2
1998 8.4 4.3 100 64.0 97.9 100.8 104.2 83.9 100.2 102.6 103.7 100.0
2003 8.3 4.8 109 57.4 105.4 87.3 97.2 118.6 94.6 113.8 95.8 85.8
South
1993 12.2 8.9 124 63.5 106.5 108.7 73.7 118.0 110.4 89.1 94.3 81.3
1998 11.3 7.3 120 68.2 99.5 114.5 113.8 82.9 108.8 91.3 79.7 101.2
2003 6.8 4.4 106 57.7 93.8 107.3 83.1 122.2 118.3 94.2 103.0 93.7
Central
1993 15.3 11.0 132 48.0 100.8 117.8 80.8 118.0 104.8 73.4 102.3 84.3
1998 9.8 6.5 114 48.3 91.9 112.4 80.2 111.3 106.0 92.2 114.7 82.7
2003 9.2 6.3 112 70.1 101.7 106.1 91.5 110.0 90.7 91.8 114.5 79.8
North
1993 17.5 11.9 137 68.0 116.5 105.6 93.2 102.9 89.8 86.5 97.1 84.5
1998 7.4 5.7 120 53.3 109.1 110.0 97.9 94.0 106.7 83.8 98.6 92.7
2003 9.0 6.3 099 41.2 103.0 100.7 102.6 88.3 101.9 97.4 93.9 102.8
East
1993 26.2 20.5 193 111.2 106.2 116.6 60.6 147.9 110.4 86.7 123.4 66.7
1998 17.5 13.7 161 66.7 105.3 110.4 77.5 132.0 88.0 81.1 106.1 76.6
2003 9.9 6.9 124 51.2 99.3 102.7 74.5 116.2 103.5 90.0 114.2 90.0
85
Table IV.1.2.4. Household Age and Sex Ratios and Myers, Bachi, Whipple and United Nations Indices for Household Data by
Region and Type of Place of Residence, TDHS 1993, 1998 and 2003 (Continued)
Myers Index
(Females)
Bachi Index
(Females)
Whipple Index
(Females)
UN
Index
Age Ratios (Females) Sex Ratios
10-14 15-19 45-49 50-54 10-14 15-19 45-49 50-54
Type of Place of Residence
Urban
1993 13.6 10.0 131 54.0 108.2 108.6 73.7 128.9 104.5 94.1 105.3 80.0
1998 8.6 4.6 107 42.5 93.3 105.2 98.9 90.7 104.6 99.7 105.4 98.5
2003 7.2 4.7 109 51.9 101.9 95.7 91.3 116.3 97.0 107.6 104.8 87.9
Rural
1993 17.8 12.6 149 56.2 107.8 113.9 82.6 115.0 102.0 81.5 97.6 76.6
1998 16.3 10.3 138 40.3 109.3 113.6 87.7 108.3 93.6 80.4 94.8 80.0
2003 9.9 6.2 115 38.4 99.6 105.2 92.5 108.3 104.9 81.8 98.2 85.3
Total
1993 15.7 11.0 138 43.7 108.1 110.8 77.0 123.1 103.4 88.9 101.9 78.4
1998 10.5 6.5 117 34.1 99.5 108.2 94.8 97.4 99.8 92.1 101.9 90.5
2003 8.2 5.2 110 39.3 101.1 98.8 91.6 113.9 99.7 98.8 102.9 87.1
86
Table IV.1.2.5. Myers, Bachi, Whipple Indices for Household Data by Demographic Characteristics of Respondent whom the
Household Interview is Completed, TDHS 1993
Myers Index (Females) Bachi Index (Females) Whipple Index (Females)
1993 1998 2003 1993 1998 2003 1993 1998 2003
Relationship with HH Head
HH Head 14.8 9.8 9.7 9.8 4.7 5.6 138 116 109
Wife/Husband 15.5 10.2 7.0 10.6 5.5 4.3 134 108 112
Son/Daughter 15.3 12.8 9.2 10.2 9.2 7.3 141 131 110
Other 15.9 14.5 10.2 12.5 11.4 8.0 149 137 108
Age
15-34 17.2 10.2 12.0 11.3 8.0 7.6 124 112 101
35-54 14.8 12.5 8.9 9.0 6.7 5.1 130 107 111
55+ 33.0 23.6 18.5 19.8 13.2 10.4 196 155 140
Sex
Male 12.6 11.8 8.3 9.4 7.2 5.0 134 120 109
Female 16.2 10.6 7.9 11.6 6.5 5.3 139 116 111
Education
No educ/Primary incomplete 24.2 19.1 12.6 15.8 10.1 7.9 164 144 118
Primary 11.0 10.1 7.2 8.4 5.8 5.0 123 110 109
Secondary + 12.2 5.4 6.5 8.4 3.9 4.6 121 104 107
Total 15.7 10.5 8.2 11.0 6.5 5.2 138 117 110
87
Figure IV.1.2.7. illustrates the percentage of women aged 10 to 60 for THDS-1993,
TDHS, 1998 and TDHS-2003. The fluctuation at the years ending with “0” and “5”
is a common problem for all three surveys. Similar to the trends at Myers, Whipple
and Bachi indices for THDSs, the quality of the data for the surveys are increasing in
terms of age heaping and digit preference. The fluctuation at THS-1993 is more vivid
and the least heaping problem is seen in TDHS-2003. It is also interesting that after
the last eligible age 49 for individual interview, the percent of the female population
is significantly increasing. The interviewers or the respondent with whom the
household list is completed may carry the eligible women to not eligible ages.
Figure IV.1.2.7. Percentage of Women 10 to 60 Years,
TDHS 1993, 1998 and 2003
0.00
0.50
1.00
1.50
2.00
2.50
3.00
10 15 20 25 30 35 40 45 50 55 60
TDHS-1993 TDHS-1998 TDHS-2003
%
Age
88
IV.1.3. Boundary Effects
The age group distortion is one of the important problems at household surveys
where eligibility is directly related with age information. Age heaping will result in
carrying an eligible women to a not eligible age easily. During the field survey
wither the interviewer herself or the respondent will make age heaping and transfer
the eligible women outside the eligible age boundaries. Table IV.1.3.1 indicates the
Boundary Effect indices based on household data at TDHS-1993, TDHS-1998 and
TDHS-2003 by region. The age and sex ratios of age groups are used for calculating
the boundary effect indices. While lower boundary effect index is estimated among
the relations with the age groups 10-14 and 15-19; upper boundary effect index is
with 45-49 and 50-54 age groups.
Table IV.1.3.1 indicates that total boundary effect is highest at TDHS-1993 with
86.9. The result indicates that the out transference of the women from the eligible
age groups is at moderate level. At TDHS-1998 and TDHS-2003, the total boundary
effect index is estimated inside low ranges (30.4 and 39.5 respectively). Although
there is no clear difference among urban and rural females at the last TDHS, at
TDHS-1998 while at urban areas the total boundary effect is estimated 181, for rural
areas it is 53.0. The lowest scores are estimated at TDHS-1998. The highest total
boundary effect score is calculated for East region at TDHS-1993 (178.1). The score
is one of the two “high” level score among all regions at the surveys. the other one is
estimated in Central region with a score of 103.5 at the same survey. The lowest
scores are estimated at the North region for all three surveys. The scores are all
inside “low” level and very close to “negligible” for the last two surveys.
The lower and upper boundary effect indexes are also presented at Table IV.1.3.1.
Results show that although for the first two surveys there are more women at the first
eligible age group than the previous one, at TDHS-2003, although the score is very
low, the possibility of the problem of transferring the eligible women from age 15-19
to age 10-14 is seen. In TDHS-1993 the only two regions which has more number of
women at age 10-14 with regard to age 15-19 is West (-16.6) and North (-7.5). There
89
is no sign of carrying eligible women to not eligible 10-14 age group is seen at other
three regions and both urban and rural populations as well. At TDHS-1993 both
among the regions and residential areas no transference of women from 15-19 age
group to 10-14 is seen. On the other hand, at TDHS-2003 except the North region, at
all regions and urban areas excess of women at age 10-14 as compared to age 15-19
is seen.
Upper Boundary effect is seen more common and vital in TDHSs. Although the level
of the problem decreased from -69.6 to -14.0 at TDHS-1993 and TDHS-1998, at the
last survey the score again estimated as -38.1. The interviewers might try to carry the
women at their last years of reproductive ages to 50-54 age group to avoid the heavy
workload of filling a complete birth history. The lowest scores which indicates high
problem are calculated for East region for all three surveys. The lowest problem is
seen, on the other hand, at North region except the results of South region at TDHS-
1998. Opposite to other region results, at South the number of women at ages 45-49
is more than 50-54 age group. This situation is also true for North region at TDHS-
2003. Except the result for urban areas at TDHS-1998, for both urban and rural areas,
the transference of women to not eligible ages is possible.
Table IV.1.3.1. Indices of Age Eligibility Distortion Based on Household Data by
Region and Type of Place of Residence, TDHS 1993, 1998 and 2003
1993 1998 2003
L* U**
Total
( |L| + |U| ) L* U**
Total
( |L| + |U| ) L* U**
Total
( |L| + |U| )
Region
West -16.6 -65.1 81.7 16.4 -14.0 30.4 -37.3 -31.4 68.7
South 23.4 -57.3 80.7 32.5 52.3 84.8 37.7 -48.4 86.1
Central 48.4 -55.1 103.5 34.3 -63.1 97.5 3.4 -53.2 56.6
North -7.5 -22.2 29.8 23.8 -2.0 25.8 2.2 23.1 25.3
East 34.0 -144.0 178.1 12.0 -84.0 96.1 16.9 -65.8 82.7
Type of Place of Residence
Urban 10.8 -80.5 91.3 16.8 1.3 18.1 -16.7 -41.9 58.6
Rural 26.7 -53.4 80.1 17.6 -35.4 53.0 28.7 -28.8 57.6
Total 17.2 -69.6 86.9 16.4 -14.0 30.4 -1.4 -38.1 39.5
* Lower boundary
** Upper Boundary
90
Table IV.1.3.1. shows the boundary effect index results according to some sociodemographic
characteristics of the respondent whom the household questionnaire is
filled with. The total boundary effect results indicate that, the household head’s
spouse gives the best results as compared to the other household members. For the
last two surveys the level of total boundary effect is even at negligible level (12.3
and 23.9 respectively). When the age of the respondent is taken into consider, the
best results are seen among 55+ ages at the first two surveys and among 35-54 age
group at TDHS-2003. The level of error at age group 55+ at TDHS-1998 is very low
(6.0) as compared to all other ages and years. The sex and the educational status of
the respondent seem not significant in terms of total boundary effect scores. For the
first and the last TDHSs, males’ score is worse than female respondents’. In addition,
the total boundary error score of respondents who have no education or did not
complete the primary school is lower for the first two surveys. However, at TDHS-
2008, the lowest score is calculated among the respondents having secondary or
more education.
The lower boundary effect results shows us that, the transference of women to 10-14
age group from 15-19 age group is seen at TDHS-2003 among the respondents who
are wife or husband of the household head. -22.8 lower boundary effect score is
calculated for this group which is the lowest score among all three years and sociodemographic
characteristics. On the contrary, mentioning more females at the first
eligible age group is more common situation among the son or daughter of the
household head for all three TDHSs. The age of the respondents seem no clear
relationship with the lower boundary score. However, the results indicate that among
males –although not very high- the problem of carrying the eligible women to not
eligible ages is seen for the early eligible ages. Female respondents, on the other
hand, lists down more women to the first eligible age group than to the 10-14 age
group. For the first two TDHSs, if the respondent’s educational level increases, the
possibility of writing down more women to age group 15-19 also increases.
However, TDHS-2003 results show that, this situation is true for only members who
completed primary education. At TDHS-2003 the level of lower boundary effect is in
negligible levels.
91
Upper boundary effect seems common at TDHS datasets. Except for some categories
of TDHS-1998, more or less for every socio-demographic characteristic of
respondents; transferring eligible women from the last eligible age group to the next
is seen clearly. The lowest problem is found among the age information given by
wife or husband of the household head and their children for all three surveys. In
addition, the quality of the responses is increasing at the advanced ages at TDHS-
1993 and TDHS-1998. At the last survey, for all age groups the upper boundary
effect scores are significantly low.
Table IV.1.3.2 also presents the upper boundary effect results by sex of the
respondent and the education level. While the score at TDHS-1993 is remarkably
low for both males and females; male responses are worse than females (-65.3 and -
48.0 respectively). The same situation is true for the last survey with better results.
TDHS-1998 have the best of the worst results in terms of upper boundary effects,
while the male respondents’ score (-2.4) is negligible, females’ score (-11.9) is
around 5 times worse than males’. On the other hand, while no clear relationship is
seen with the education level and the level of upper boundary score at TDHS-1993,
for the last two surveys it is seen that, when the level of education of the respondent
is increasing the quality of the data is also increasing.
92
Table IV.1.3.2. Indices of Age Eligibility Distortion Based on Household Data by
Demographic Characteristics of Respondent whom the Household Interview is
Completed, TDHS 1993, 1998 and 2003
1993 1998 2003
L* U** Total
( |L| + |U| )
L* U** Total
( |L| + |U| )
L* U** Total
( |L| + |U| )
Relationship with
HH Head
HH Head 8.4 -60.5 68.9 -8.2 -13.6 21.8 5.4 -60.7 66.1
Wife/Husband -3.2 -40.0 43.2 7.9 4.4 12.3 -22.8 -1.1 23.9
Son/Daughter 47.3 -40.8 88.2 48.3 14.6 62.9 70.9 -9.4 80.3
Other 44.5 -64.1 108.5 43.4 -25.2 68.5 17.6 -24.4 42.0
Age
15-34 4.2 -97.1 101.4 38.8 -34.7 73.5 14.1 -33.8 47.9
35-54 25.6 -60.3 85.9 18.9 -4.9 23.8 -3.4 -21.6 25.0
55+ 47.4 -15.5 62.9 -3.6 2.4 6.0 36.4 -26.8 63.2
Sex
Male -2.4 -65.3 67.7 -1.3 -2.4 3.7 -2.6 -41.4 43.9
Female 13.7 -48.0 61.6 21.4 -11.9 33.3 5.2 -16.0 21.2
Education
No educ/Pri. Inc. 8.2 -35.6 43.9 9.6 -20.1 29.7 -6.9 -39.9 46.7
Primary 7.3 -70.5 77.7 16.5 -13.6 30.1 10.6 -18.9 29.4
Secondary + 32.0 -45.6 77.6 28.5 20.1 48.6 -2.6 -6.6 9.2
Total 17.2 -69.6 86.9 16.4 -14.0 30.4 -1.4 -38.1 39.5
* Lower boundary
** Upper Boundary
IV.1.4. The Household Residency
The household questionnaire consists a question for the members of the household
and the visitor’s status of sleeping at the dwelling the night before the survey. To
become eligible at TDHSs, a woman at ages 15-49 must be either a usual resident of
the household or a visitor who slept last night of the interview day. If there is large
difference between the number of overnight visitors and the usual residents sleep
93
away, an exclusion of eligible women problem will be brought into manner. TDHS
household data is evaluated to assess the sleeping away exclusion of female usual
residents from the household list. Table IV.1.4.1. illustrates the number of eligible
female population, percent of usual residents sleeping away at the interview date and
the percent of the overnight visitors at TDHS-1993. No systematic exclusion of usual
female members at eligible ages is seen in general except the age group 45-49. While
7.5 % of the usual female residents were mentioned as slept away, visitors only 4.5
percent of the usual residents were written down at the list at this age group.
Regional differences for the last eligible age groups are seen. While there is no
exclusion at North region, 9 percent of the usual resident female population at 45-49
age group is estimated as excluded. In addition, 2.1 %, 2.2 %, and 3.5 % of females
at this age group is assessed as excluded at West, South and Central regions
respectively. On the other hand, at the urban areas, the percent of the excluded
women are higher than rural in general. At the age group 45-49, 4.5% of the eligible
usual resident is estimated as excluded.
TDHS-1998 data is evaluated in terms of sleeping away exclusion of eligible women
at Table IV.1.4.2. No clear omission of eligible women is seen at TDHS-1998 in
general. However at Region East, for the first and the last eligible age groups, 4.4 %
and 4.7% exclusion is estimated respectively. In addition, at South region, 3.7 % of
the females aged 15-19 and 45-49 is assessed being excluded. On the other hand,
Table IV.1.4.3. shows the sleeping away exclusion of the female usual residents from
the household list of TDHS-2003. In general the omission is seen at 15-19 age group.
4.4 % of the usual resident women are excluded at this age group. 7.5 % of the rural
females and 2.7 % of urban females at this age group are also excluded. For all
regions, more or less, exclusion is calculated. The highest exclusion is seen at North
region where nearly 1 out of 10 women are excluded. The omission of women at the
last eligible age group is noticeable in rural areas and in East, North and South
regions.
94
Table IV.1.4.1. Evaluation of the Sleeping Away Exclusion of the Female Usual
Residents, TDHS-1993.
Age group
10-14 15-19 20-24 25- 30- 35-39 40-44 45-49 50-54
Region
West
A 688 665 604 514 505 452 379 289 317
B 3.9 5.5 6.9 3.8 5.8 4.1 4.2 6.8 4.6
C 3.6 5.5 5.1 3.8 5.6 4.6 5.2 4.7 7.3
Est. % excluded 0.4 0.0 1.8 0.0 0.2 -0.5 -1.0 2.1 -2.7
South
A 370 379 316 243 210 182 155 107 134
B 5.5 7.6 6.7 4.5 4.4 6.0 6.0 10.1 11.0
C 4.8 7.0 9.4 6.4 7.8 6.8 7.0 8.0 6.4
Est. % excluded 0.6 0.6 -2.7 -1.9 -3.3 -0.9 -1.0 2.2 4.6
Central
A 518 553 412 336 313 251 239 183 208
B 3.7 5.5 6.4 3.1 3.0 5.8 5.7 6.3 5.1
C 4.0 5.5 8.1 6.2 4.6 5.0 1.3 2.8 6.0
Est. % excluded -0.4 0.0 -1.8 -3.1 -1.6 0.9 4.4 3.5 -0.9
North
A 231 203 161 138 118 115 75 67 73
B 8.5 7.0 11.4 9.9 4.5 7.7 9.4 6.1 12.9
C 6.7 8.5 14.3 15.9 12.1 7.7 10.2 10.5 8.9
Est. % excluded 1.8 -1.5 -2.9 -6.0 -7.5 0.0 -0.8 -4.4 4.0
EastA 586 560 365 215 234 159 146 103 165
B 1.4 4.1 4.4 4.1 6.1 7.1 2.9 10.0 3.1
C 3.5 4.4 5.3 6.2 4.8 5.1 4.3 1.0 1.8
Est. % excluded -2.1 -0.3 -0.9 -2.0 1.3 2.0 -1.4 9.0 1.3
Type of Place of Residence
Urban
A 1389 1376 1144 972 925 755 661 453 524
B 4.4 5.8 5.7 4.1 3.9 5.9 5.4 7.5 6.9
C 3.2 5.8 7.2 4.3 4.6 4.7 4.1 3.2 7.0
Est. % excluded 1.2 0.1 -1.5 -0.2 -0.7 1.2 1.3 4.2 -0.1
Rural
A 1003 985 715 474 454 404 334 296 374
B 3,2 5,4 8,1 5,0 6,9 4,9 4,4 7,7 5,0
C 5,5 5,7 7,4 10,4 9,2 6,7 6,2 7,0 4,6
Est. % excluded -2,2 -0,3 0,6 -5,4 -2,3 -1,8 -1,8 0,7 0,4
Total
A 2392 2361 1859 1446 1379 1159 995 749 898
B 3.9 5.6 6.6 4.4 4.9 5.5 5.0 7.5 6.1
C 4.1 5.7 7.3 6.3 6.1 5.4 4.8 4.7 6.0
Est. % excluded -0.2 -0.1 -0.7 -1.9 -1.2 0.1 0.3 2.8 0.1
(A) The Number of Women reported to reside in Interviewed Households,
(B) the Percentage of Resident Women Not Sleeping In the Household During the Night Before the Survey,
(C) the Standardized Percentage of Non-Resident Women Sleeping In the Household during the Night Before the
Survey
Est. % excluded: Estimated Percentage of Women Who Were Excluded from Eligibility for the Individual
Interview by Region and Type of Place of Residence.
95
Table IV.1.4.2. Evaluation of the Sleeping Away Exclusion of the Female Usual
Residents, TDHS 1998.
Age group
10-14 15-19 20-24 25- 30- 35-39 40-44 45-49 50-54
Region
West
A 556 622 672 573 535 443 387 320 254
B 8.8 9.2 11.2 8.4 6.5 8.8 8.7 5.5 9.8
C 8.3 6.3 7.2 7.2 4.5 5.2 4.9 6.3 5.4
Est. % excluded 0.5 2.9 4.0 1.2 1.9 3.6 3.8 -0.7 4.4
South
A 265 301 229 214 173 173 138 131 89
B 3.7 8.0 4.0 8.4 1.2 1.7 3.6 6.3 9.2
C 6.2 4.3 6.3 7.1 4.9 3.6 5.4 2.6 4.5
Est. % excluded -2.6 3.7 -2.3 1.3 -3.6 -2.0 -1.9 3.7 4.6
Central
A 365 409 356 344 258 282 247 174 170
B 4.7 6.4 4.7 2.7 2.2 5.7 4.2 7.1 4.7
C 4.5 9.4 12.5 7.6 7.7 7.0 5.5 5.2 6.6
Est. % excluded 0.2 -3.0 -7.8 -4.9 -5.5 -1.3 -1.3 1.9 -1.8
North
A 150 147 117 107 105 99 84 67 58
B 10.0 10.3 14.6 10.6 5.2 7.3 7.8 5.6 8.2
C 10.9 10.9 16.0 16.5 13.3 5.2 11.4 9.9 4.5
Est. % excluded -0.9 -0.6 -1.5 -5.9 -8.2 2.1 -3.5 -4.3 3.7
EastA 496 464 318 215 164 169 130 103 131
B 6.5 9.6 8.3 4.9 6.4 6.7 8.1 9.6 5.2
C 3.7 5.2 8.7 12.7 7.6 6.6 3.5 4.9 3.3
Est. % excluded 2.8 4.4 -0.4 -7.7 -1.3 0.1 4.6 4.7 1.9
Type of Place of Residence
Urban
A 1086 1193 1163 1035 876 809 676 525 400
B 7.9 8.8 9.1 6.1 4.7 7.5 7.4 6.3 8.7
C 4.9 5.8 8.0 6.4 4.3 3.1 3.8 4.6 6.1
Est. % excluded 3.0 3.0 1.0 -0.3 0.4 4.4 3.6 1.7 2.5
Rural
A 746 750 528 418 359 356 311 269 302
B 5,0 8,3 7,3 8,1 4,7 4,4 5,2 6,9 6,0
C 8,1 8,2 11,4 14,5 11,5 11,4 9,3 7,4 3,7
Est. % excluded -3,2 0,1 -4,1 -6,4 -6,8 -7,0 -4,1 -0,5 2,3
Total
A 1832 1944 1691 1453 1235 1165 987 794 701
B 6.7 8.6 8.5 6.7 4.7 6.6 6.7 6.5 7.5
C 6.2 6.7 9.1 8.8 6.4 5.6 5.5 5.6 5.1
Est. % excluded 0.5 1.9 -0.6 -2.1 -1.7 0.9 1.2 1.0 2.4
((A) The Number of Women reported to reside in Interviewed Households,
(B) the Percentage of Resident Women Not Sleeping In the Household During the Night Before the Survey,
(C) the Standardized Percentage of Non-Resident Women Sleeping In the Household during the Night Before the
Survey
Est. % excluded: Estimated Percentage of Women Who Were Excluded from Eligibility for the Individual
Interview by Region and Type of Place of Residence.
96
Table IV.1.4.3. Evaluation of the Sleeping Away Exclusion of the Female Usual
Residents, TDHS 2003.
Age group
10-14 15-19 20-24 25- 30- 35-39 40-44 45-49 50-54
Region
West
A 723 695 787 768 653 595 567 497 467
B 2.0 9.8 5.7 3.6 2.1 2.2 1.7 3.2 5.3
C 1.3 4.7 6.1 8.0 2.9 2.7 3.2 5.3 5.1
Est. % excluded 0.7 5.2 -0.4 -4.5 -0.8 -0.5 -1.5 -2.1 0.2
South
A 283 306 291 246 232 207 175 138 145
B 3.1 9.3 12.1 6.7 2.4 2.2 2.6 4.5 5.2
C 1.4 5.3 4.2 4.3 2.8 2.6 3.7 1.4 4.5
Est. % excluded 1.7 4.0 7.9 2.3 -0.3 -0.4 -1.1 3.1 0.8
Central
A 454 511 493 378 344 361 360 275 236
B 2.6 6.5 6.2 3.4 1.2 3.2 4.5 2.8 3.9
C 2.5 3.9 4.1 1.9 2.8 1.5 2.6 2.6 9.0
Est. % excluded 0.2 2.6 2.1 1.5 -1.5 1.7 1.9 0.1 -5.1
North
A 162 169 154 118 136 112 123 105 74
B 6.5 12.7 8.3 3.9 2.8 2.8 1.9 7.9 4.9
C 3.4 3.5 4.0 5.5 2.0 3.4 0.0 2.2 2.8
Est. % excluded 3.1 9.2 4.3 -1.6 0.8 -0.6 1.9 5.7 2.2
EastA 531 520 453 346 271 238 184 127 141
B 3.4 6.7 6.5 4.9 2.0 4.1 5.2 7.6 4.5
C 1.2 2.8 5.7 3.8 2.6 2.3 3.9 2.0 3.7
Est. % excluded 2.3 3.9 0.8 1.1 -0.6 1.9 1.3 5.6 0.8
Type of Place of Residence
Urban
A 1417 1423 1500 1389 1154 1058 1014 791 719
B 2.0 7.7 7.3 3.8 1.6 3.0 2.7 3.2 5.1
C 2.0 4.9 5.7 5.5 3.1 2.5 3.1 4.2 6.7
Est. % excluded -0.1 2.7 1.5 -1.8 -1.5 0.5 -0.4 -1.0 -1.5
Rural
A 737 778 678 466 481 456 396 352 737
B 5,0 10,0 6,6 5,6 3,0 2,2 3,9 6,3 5,0
C 1,0 2,4 4,0 4,8 1,9 2,2 2,5 2,1 1,0
Est. % excluded 3,9 7,5 2,6 0,8 1,1 0,0 1,4 4,2 3,9
Total
A 2154 2201 2177 1855 1636 1514 1410 1142 1063
B 3.0 8.5 7.0 4.2 2.0 2.8 3.0 4.2 4.8
C 1.7 4.0 5.2 5.3 2.7 2.4 2.9 3.5 5.5
Est. % excluded 1.3 4.4 1.9 -1.1 -0.7 0.4 0.1 0.6 -0.7
(A) The Number of Women reported to reside in Interviewed Households,
(B) the Percentage of Resident Women Not Sleeping In the Household During the Night Before the Survey,
(C) the Standardized Percentage of Non-Resident Women Sleeping In the Household during the Night Before the
Survey
Est. % excluded: Estimated Percentage of Women Who Were Excluded from Eligibility for the Individual
Interview by Region and Type of Place of Residence.
97
IV.2. The Assessment of Age Data in Individual Questionnaire
The household questionnaire -in its nature- as filled with a proxy respondent, is more
vulnerable to age reporting errors. Besides the quality of the age and sex data used
for the decision of the eligibility of the individual interview from the household
questionnaire, the quality of the age data collected by the individual questionnaire is
very important in terms of estimating fertility rates. Similar methods and indexes
used for age at household list are used to assess the data quality of age information at
individual data set. The age information at individual questionnaire is supported by
the year and month questions with respect to household questionnaire. It both gives
the opportunity for the individual and respondent to make a connection with the birth
date and the age of the respondent at time of survey. A woman will be sure about her
birth date and either she or the respondent or both may calculate her birth date with
this information. Another situation is woman can know her age and respondent will
probe her birth date with the age information given and make women confirm the
birth year. At this section of the study, the women’s age information is assessed from
different points.
IV.2.1. Digit Preference
Digit preference seems one of the common problems in surveys and censuses
especially for the developing and undeveloped countries. Table IV.2.1.1 presents the
percent distribution of women 20-49 by reported terminal digit of age and Myers
Index by Region, Type of Place of Residence and Education of Woman. The
percents of the digits mentioned for ages indicates that, there is noticeable digit
preference in general for the digits “0”, “3”, “5” and “8”. “0” and “5” are the
“universal” digits preferred by the individuals. In addition, TDHS surveys are
applied at the years ending with 3 and 8. Rounding the birth years to the years ending
with “0” and “5” will result in digit preference for “3”and “8”. For the East Region a
tendency to prefer the digits “0”, “3” and “5” is vivid. Although the preference of
these digits decreased with the last survey, a clear preference is seen.
98
In the rural areas, on the other hand, a tendency to heap the age to the years ending
with these three digits is also remarkable. Preferences of these digits are seen for all
three surveys. In urban, only digit having a percent more than 10 for all three surveys
is “3”.
Digit preference is also evaluated with the education of the respondent. Except the
results of TDHS-1998, when the level of education increases, the digit preference for
“0” is decreasing. For other digits, it is difficult to mention a relation between the
education of the woman and the digit preference.
Myers index is assessed for women 20-49 and also presented at Table IV.2.1.1. It is
seen that the index results are lower at individual questionnaire as compared to the
household for all surveys. For the first and the last TDHSs, the index result is
remarkably low. It is also seen that the Myers index results at TDHS-2003 is only
4.0. When the results are evaluated in terms of regional differences, it is seen that
For the first two surveys the Myers index results are highest at East Region. The only
region at where the Myers Index result is higher at TDHS-2003 as compared to
TDHS-1993 is North. At this region Myers index result increased gradually from
10.6 to12.9 and then 15.2 among the three survey chronologically. Although there
are fluctuations at the index results, except South for all regions the clear
improvement at the data quality of the age mentioned is seen.
Myers index results are at least 2 point lower in urban areas to rural at all TDHSs. In
urban areas, the index drops nearly 7 points and in rural 6 points from TDHS-1998 to
TDHS-2003. Although there is a fluctuation for rural areas, a gradual decrease is
seen at urban. As mentioned above for the digit preference, a positive correlation
between the education of woman and the index result is not seen. While the Myers
index results decreases between the first and the last survey (ignoring TDHS-1998),
among the women who had secondary or more education there is around 3 point
increase. TDHS-1998 results are very dislike the other two survey results and create
a fluctuation at Myers index results.
99
Table IV.2.1.1. Percent Distribution of Women 20-49 by Reported Terminal
Digit of Age (Individual Questionnaire) and Myers index by Region, Type of
Place of Residence and Education of Woman, TDHS 1993, 1998 and 2003
Terminal Digit Myers
0 1 2 3 4 5 6 7 8 9 Index
Region
West
1993 9.7 8.8 10.6 11.4 10.8 10.3 9.3 10.2 10.8 8.1 8.2
1998 11.3 8.7 10.9 11.3 11.5 8.3 10.5 9.7 10.0 7.8 11.0
2003 10.1 8.9 9.8 9.8 9.2 10.3 10.1 10.8 10.8 10.2 4.6
South
1993 10.3 9.6 8.8 11.4 11.0 11.6 8.0 9.7 11.9 7.7 12.4
1998 11.9 9.2 9.7 11.0 10.1 11.8 10.9 9.0 9.6 6.7 11.5
2003 9.1 8.8 10.5 11.7 10.2 10.0 9.8 10.3 9.5 10.0 5.5
Central
1993 10.5 10.1 9.3 12.4 10.4 10.9 8.9 9.0 10.3 8.3 9.1
1998 10.8 10.1 9.3 11.3 10.7 9.1 10.6 8.4 11.2 8.4 9.5
2003 9.8 10.0 11.1 8.8 9.9 11.1 9.1 8.7 11.7 9.8 7.8
North
1993 11.4 9.4 9.6 9.4 8.0 9.0 9.6 11.3 12.5 9.6 10.6
1998 12.8 8.6 9.9 10.8 11.4 11.5 9.4 7.5 9.6 8.6 12.9
2003 8.9 8.0 8.7 11.9 11.6 8.3 10.0 8.4 13.5 10.5 15.2
East
1993 19.0 8.8 9.2 10.0 9.7 12.6 6.4 8.9 9.1 6.3 23.2
1998 16.7 7.5 10.2 11.4 8.8 13.6 9.6 8.9 8.5 4.9 23.7
2003 12.0 8.0 11.3 10.9 9.8 12.0 8.7 8.5 9.6 9.3 12.3
Type of Place of Residence
Urban
1993 10.8 9.0 9.9 11.4 11.4 10.8 8.8 9.7 10.7 7.6 10.1
1998 11.7 9.1 10.4 11.4 10.7 9.6 10.2 9.8 9.9 7.3 8.7
2003 10.0 8.9 10.1 10.1 9.9 10.6 9.7 10.0 11.0 9.8 3.5
Rural
1993 13.2 9.7 9.5 11.0 8.3 11.1 8.2 9.8 10.9 8.4 12.3
1998 13.3 8.4 9.8 10.9 10.6 11.1 10.8 7.4 10.0 7.8 13.3
2003 10.5 9.0 10.9 10.3 9.4 10.6 9.4 9.0 10.5 10.4 6.4
100
Table IV.2.1.1. Percent Distribution of Women 20-49 by Reported Terminal
Digit of Age (Individual Questionnaire) and Myers index by Region, Type of
Place of Residence and Education of Woman, TDHS 1993, 1998 and 2003
(Continued)
Terminal Digit Myers
0 1 2 3 4 5 6 7 8 9 Index
Education
No educ/Pri. Inc.
1993 13.4 6.8 8.4 10.6 9.1 13.1 8.5 10.3 11.7 8.0 18.3
1998 12.7 4.8 8.7 10.7 9.5 13.4 9.7 9.7 13.2 7.4 20.2
2003 10.9 7.0 10.7 10.1 9.3 10.1 8.1 9.9 12.2 11.7 11.4
Primary
1993 11.1 10.9 10.4 11.5 10.5 9.7 8.7 9.4 10.5 7.4 9.7
1998 11.1 9.9 10.0 11.6 11.2 9.3 11.6 8.5 9.4 7.4 11.0
2003 10.3 9.8 10.9 9.9 9.5 10.5 9.3 9.1 11.1 9.6 5.6
Secondary +
1993 8.4 8.9 10.6 11.8 13.0 9.7 8.5 9.3 9.0 10.8 12.4
1998 14.7 11.2 12.7 10.9 10.5 8.0 7.5 9.7 7.4 7.5 19.9
2003 8.6 8.4 7.9 11.0 11.2 11.5 12.4 11.8 8.2 9.0 15.8
Total
1993 11.6 9.3 9.7 11.2 10.3 10.9 8.6 9.7 10.8 7.9 9.6
1998 12.2 8.9 10.2 11.2 10.7 10.1 10.4 9.0 10.0 7.4 9.5
2003 10.1 8.9 10.3 10.2 9.8 10.6 9.6 9.7 10.8 10.0 4.0
IV.2.2. Imputation at the Age Data
Women who have no connection to the registration system may have problem in
remembering her birth month and year and even her age. At the countries where the
registration “culture” is not placed, people lives problem of the dates of the vital
events in their life. As mentioned before TDHS individual questionnaire includes
two questions to get the exact age of the women which is very valuable in estimation
of rates, especially fertility rates. Both the birth month and year and current age of
the woman is asked by the interviewer. This information is both checked by the
interviewer and field editor at the field and during the data entry by the entry
101
program for internal consistency. Defective and inconsistent data collected from the
field is corrected during data entry with using the standard imputation procedures,
partial information will be completed after the editing procedure at the data entry
when needed.
Table IV.2.2.1 indicates completeness of the date of birth and age information by
region, type of place of residence. Except TDHS-1998 for the other two surveys, the
around 80 % of the data collected on birth month and year and age is completed at
the field. In THSDS-198 this is true for 71.4 % of the interviews. Around 20 % of the
interviews, age is given by the respondent and year is calculated with imputation of
the month in the same survey. At TDHS-1993 and TDHS-2003 this occurred at 16.0
% and 13.5 % of the interviews respectively. Clear regional difference is seen
especially between West and East. The highest complete information percent is
collected at West region with 89.0 %, 82.7% and 88.4 % at TDHS-1993, TDHS-
1998 and TDHS-2003 in that order. On the other hand, at East region the lowest
percents are seen among all regions for all three surveys in terms of completeness of
the age and birth date information. Only 68.4 % of the responses were complete and
this dropped to 56.4 % at TDHS-2003. It is very surprising that less than half of the
age and birth date information is complete at TDHS-1998. The gap between East
region and the rest is opened in connection with completeness of the data with
TDHS-2003.
The respondents living in urban areas seem good at remembering their birth date and
age together as compared to the ones living in rural residences. At least 10%
difference is seen between urban and rural responses regarding the completeness of
the birth date and age data. Although TDHS-1998 results disrupt the trend, for both
urban and rural responses the percent of completeness decreases.
The common situation where the completeness percents are low is the one at which
respondents give the age information and with imputing the month a year value is
assessed. In TDHS-1993 nearly one third of the respondents mentioned their age
correctly and month is imputed and a year is assigned for these questionnaires at East
102
region. As both this situation and completeness rate decreases, the condition at which
only year is mentioned and month is imputed and age is calculated increases. This is
clearly seen at East region and rural areas. Another eye catching result is on the ,east
region where at TDHS-1993, only 1.3 percent of the respondents give their year
information and age is calculated by imputing the month information. However, this
is seen at 22.7 % of the interviews for the last two TDHSs which is very surprising.
On the other hand, results indicate that, at the South region, while the completeness
percent of the age and birth date data is decreasing from TDHS-1993 to TDHS-2003,
the situation where the age information is given and year is calculated with the
imputation of the month information.
103
Table IV.2.2.1. Percent Distribution of the Completeness of the Date of Birth
and Age Information by Region and Type of Place of Residence, TDHS 1993,
1998 and 2003
Given Year & Month Year & Age Age Year
Imputed Month Month Month Total
Calculated Age Year Age % n
Region
West
1993 89.0 0.2 10.1 0.6 100.0 2325
1998 82.7 0.2 12.9 4.2 100.0 3204
2003 88.4 0.2 9.6 1.8 100.0 3286
South
1993 83.6 0.2 15.8 0.5 100.0 998
1998 68.0 0.0 24.2 7.8 100.0 1258
2003 72.0 0.5 21.7 5.8 100.0 1028
Central
1993 85.5 0.1 13.5 0.9 100.0 1520
1998 71.1 0.8 19.2 9.0 100.0 1985
2003 85.8 0.3 12.5 1.4 100.0 1867
North
1993 75.0 0.2 20.1 4.7 100.0 612
1998 73.0 0.2 22.1 4.6 100.0 692
2003 80.8 0.7 14.5 4.0 100.0 590
East
1993 68.4 0.2 30.2 1.3 100.0 1064
1998 48.6 0.5 28.2 22.7 100.0 1437
2003 56.4 3.2 17.7 22.7 100.0 1305
Type of Place of
Residence
Urban
1993 87.5 0.2 11.7 0.7 100.0 4181
1998 77.8 0.2 15.5 6.4 100.0 5704
2003 84.7 0.7 10.7 4.0 100.0 5752
Rural
1993 74.2 0.2 23.6 2.0 100.0 2338
1998 58.6 0.5 26.9 14.0 100.0 2872
2003 68.3 1.1 20.4 10.2 100.0 2323
Total
1993 82.7 0.2 16.0 1.2 100.0 6519
1998 71.4 0.3 19.3 9.0 100.0 8576
2003 80.0 0.8 13.5 5.8 100.0 8075
104
Table IV.2.2.2. demonstrates the completeness percent of the date of birth and age
information by age and education of the respondent. The highest completeness rate is
gathered by the women under 30. At TDHS-1993 and TDHS-1998, 87.1% and
83.9% of the respondents aged 15-19 gave complete information respectively which
is highest for these surveys. At TDHS-2003 the highest rate is seen among 20-2 age
group of women with 87%. Lowest completeness rates, on the other hand, are seen
among the last eligible age group. Difference between 15-19 and 45-49 age groups
are, 10.5 %, 28.4% and 7.5% at TDHS-1993, TDHS, 1998 and TDHS-2003 in that
order. The gap at the first and the last eligible age groups in terms of complete age
and birth date information is remarkably high at TDHS-1998.
Education of women seems have positive relation on the quality of the age and birth
date data. When the education level of the women increases the completeness of the
data is getting close to perfect. Nearly all of the women having secondary or more
education gave complete age and birth date information. On the contrary, women
have no education or did not complete the primary school gives the highest
incomplete information. While TDHS-1993 individual data shows that, two thirds of
women at the lowest education level gave complete information, completeness rates
dropped to 30.3 % at TDHS-1998 and 44.2% at TDHS-2003.
105
Table IV.2.2.2. Percent Distribution of the Completeness of the Date of Birth
and Age Information by Demographic Characteristics of Women, TDHS 1993,
1998 and 2003
Given Year & Month Year & Age Age Year
Imputed Month Month Month Total
Calculated Age Year Age % n
Age
15-19
1993 87.1 0.2 11.2 1.5 100.0 332
1998 83.9 0.4 10.3 5.4 100.0 1720
2003 81.5 0.9 7.5 10.1 100.0 238
20-24
1993 86.4 0.1 12.8 0.7 100.0 1040
1998 80.2 0.1 13.5 6.2 100.0 1558
2003 87.0 1.0 7.4 4.6 100.0 1045
25-29
1993 85.8 0.3 12.9 0.9 100.0 1211
1998 73.9 0.8 17.0 8.3 100.0 1397
2003 85.6 0.7 8.8 4.9 100.0 1480
30-34
1993 83.9 0.0 15.3 0.8 100.0 1283
1998 68.4 0.1 21.6 9.9 100.0 1202
2003 80.0 0.9 13.8 5.4 100.0 1489
35-39
1993 80.2 0.3 18.1 1.4 100.0 1073
1998 61.9 0.5 26.7 10.9 100.0 1081
2003 77.0 0.8 16.2 6.1 100.0 1420
40-44
1993 78.3 0.1 20.0 1.6 100.0 901
1998 56.1 0.2 30.7 13.0 100.0 885
2003 75.9 0.8 16.5 6.7 100.0 1330
45-49
1993 76.6 0.3 21.4 1.7 100.0 679
1998 55.5 0.3 28.9 15.3 100.0 733
2003 74.0 0.7 19.3 6.0 100.0 1073
Education
No educ/Pri. Inc.
1993 67.7 0.2 29.2 2.9 100.0 2196
1998 30.3 0.5 37.3 31.8 100.0 1861
2003 44.2 3.0 30.0 22.7 100.0 1761
Primary
1993 88.6 0.2 10.9 0.3 100.0 3662
1998 77.8 0.3 18.5 3.4 100.0 5158
2003 87.4 0.2 11.1 1.3 100.0 4940
Secondary
1993 99.7 0.0 0.3 0.0 100.0 661
1998 99.2 0.1 0.5 0.2 100.0 1556
2003 99.2 0.1 0.7 0.0 100.0 1374
Total
1993 82.7 0.2 16.0 1.2 100.0 6519
1998 71.4 0.3 19.3 9.0 100.0 8576
2003 80.0 0.8 13.5 5.8 100.0 8075
106
The assessment of data used to determine the eligibility for the individual interview
is crucial in terms of the number of ever married women added for all means of rates
and ratios calculated from the TDHS. The quality of the data at household
questionnaire is assessed at this chapter of the study to put the potential problems of
the data into the matter of discussion. Chapter starts with the household interview
results which gives an overview of the result codes of the responses. It is seen that
the completed questionnaires are lower at TDHS-1998 as compared to the first and
the last TDHS. On the other hand, the response rates are decreasing during the
surveys. The response rates even decreased to 90 % at region West which will bring
questions to mind of the users of the data about the respresentativeness of it at this
region.
The quality of the age reporing in household questionnaire is assessed at this chapter.
The age distribution of the de facto population is discussed in terms of heaping and
digit preference for males and females separately. It is seen that, although the
heaping is seen at both sexes the level of heaping is high at females. At this section
of the chapter, the Myers, Whipple, Bachi and UN Age-Sex Ratio indices were used
to estimate the problems of age at household data. In addition to the nearly universal
digits preferred 0 and 5; at TDHSs 3 and 8 is also preferrered by the respondents at
the survey years are ending with 3 and 8.
The quality of the age data of the female members of the household is at the center of
the section. Age and sex ratios and Myers, Bachi, Whipple and United Nations
indices for household data for total, regions residencial difference is estimated for
female members. The index results for females show that the age heaping problem is
decreasing at TDHSs. While the results of the first two TDHSs are evaluated as
medium quality, the age distribution of TDHS-2003 female member’s age
distribution is low and acceptable. A gradual increase at the data quality of the age
information is seen at both urban and rural females. In addition same indices and
indexes are done for the selected characteristics of the respondent whom the
household questionnaire is completed with. The results indicate that the best
information is taken from either household head or her/his spouse.
107
The upper, lower and total boundary effect problems which are mainly sourced from
the displacement of eligible women out of eligible ages are evaluated for all TDHSs
at this chapter. While there seems no problem at the lower boundary, at upper
boundary where the women at the end of reproductive ages are carried to age group
50-54 is quite common. While the level of the problem decreased from TDHS-1993
to TDHS-1998, at the last survey the upper boundary effect seems increased. The
result that indicates that the highest problem is seen at East region is important as the
TFR at this region is high and exclusion of women may lead problems at this region.
The charactersistics of the respondent is also taken into consider while estimating the
boundary effects. Results show that the household head’s spouse gives the less
problematic age information in terms of boundary effects.
The sleeping away exclusion of the women from the household list is also evaluated
with the results of the household residency answers at the household list. The
difference between the number of overnight visitors and the usual residents sleep
away is studied in terms of finding the sleeping away exclusion of women. It is seen
that except the last eligible age group no significant sleeping away exclusion is seen
at TDHS-1993. At TDHS-1998, on the other hand at Region East omission of
eligible women at the first and the last eligible age group is seen caused by sleeping
away exclusion. In general, the omission is seen at 15-19 age group where the usual
resident women are excluded at this age group at the last TDHS.
The assessment of the age data in individual questionnaire is studied at the second
section of this chapter. Similar techniques and indices are used to evaluate the age
data of the women at individual questionnaire. The digit preference assessment is the
first part of this section. Myers index and the percent distribution of the ages of 20-
49 women are estimated for the assessment. Like the household questionnaire, the
percents of the digits mentioned for ages indicates that, there is noticeable digit
preference in general for “0”, “3”, “5” and “8”. At some regions although the
preference of these digits decreased at TDHS-2003, the digit preference is more clear
especially “0” “3” and “5”. In addition, results show that except TDHS-1998, the
education of the women has positive effect on avoiding from digit preference.
108
The completeness of the birth month, year and age is also studied at this section of
the chapter. The level of imputation at the age data is also one of the cruacial study
points at this section. The completeness of the age information is lower at TDHS-
1998 as compare to the first and the last survey. Around10 % difference is seen at
this survey with regards to other two. The data collected from the respondents living
in urban areas seem good as compared to the ones living in rural residences. Women
aged under 30 gives more complete birth date data than the women 30 or older. On
the other hand, it is seen that when the education level of the women increases the
completeness of the data is getting close to perfect. Nearly all of the women having
secondary or more education gave complete age and birth date information at all the
three TDHSs.
109
V. THE ASSESSMENT OF THE QUALITY OF THE BIRTH HISTORY
DATA
V.1. The Quality of Birth Related Data
One of the aims of DHS program is to collect reliable and comparable fertility and
infant and child mortality data all over the world, especially at the undeveloped and
developing countries. In addition to the quality of the age data, the quality of the
birth history data has direct relation with these rates and ratios of fertility and
mortality. Hence, to collect data with high quality is aimed at TDHSs. At this
section of the study, the birth history data of TDHS-1993, TDHS-1998 and TDHS-
2003 is assessed. In addition to the evaluation of completeness of the information on
birth dates of the children, the displacement of the birth dates, the age heaping
problem, the miscalculation of year of birth and the coverage of live births is also
assessed at this section for a complete appraisal for the birth history data in terms of
the information used for the fertility estimations.
V.1.1. Completeness of the information of Birth Dates of the Children
Birth history section takes place at “Section 2. Reproduction” at the ever married
woman questionnaire. After Brass type questions about the reproductive history of
the woman; detailed information for each live birth is collected. Section starts with
asking the mother the birth date of the child. Table V.1.1.1. illustrates the percent
distribution of children born by completeness of information on date of birth and
current age of children by region and type of place of residence. The overall results
show that, the percent of the not imputed birth date information of the children data
decreases among the first and the last survey. While 96.4 % of the birth dates were
110
not imputed at TDHS-1993, the non-imputed cases are 93.0 % at TDHS-2003. Like
the previous estimations done other sections of this chapter, TDHS-1998 results
seem to break down the trend among the tree surveys. The percent of the nonimputed
information is lower at all analysis groups at TDHS-1998 when compared to
other two surveys. 85.5% of the total births at TDHS-1998 are not imputed in terms
of birth dates of the children at birth history section.
The information collected from the women living in urban areas is in better quality
than the data collected in urban areas. While only 2.6 percent of the birth date
information gathered at urban areas is imputed in TDHS-1993, for the same year
percent of the births imputed is 4.9. The difference at urban and rural data increases
to 4.7% at TDHS-2003. There is no clear difference among the regions in terms of
the completeness of the birth date information at birth history section. For all regions
at TDHS-1993, the percent of the births with no information is above 90%. The
highest and lowest completeness rates are seen at West and North regions for this
survey respectively. At TDHS-1998, West (90.5 %) and East (73.0 %) are the
regions where the highest and lowest completeness is seen. At TDHS-2003, the
women living at Central region, on the other hand, with a percent of 96.7 mentioned
complete information of their children’s birth dates with the highest completeness
rate whereas at East region 84.3 % of the birth date is information is respondent
completely.
Reporting the year and age and imputation of the month is the commonly seen
situation where the imputation is done. At TDHS-1998, the highest percent for this
situation is seen 9.5 % of the responses. Mentioning year and age and the imputation
of month is higher in rural areas for all surveys (respectively, 2.5 %, 11.2 % and 5.8
%) and at East Region at TDHS-1998 (19.5 %) and TDHS-2003 (8.5 %). At North
region, imputation of only month data is seen highest at TDHS-1998 with a percent
of 3.9.
The cases where month and age reported but year imputed; age reported but year and
month imputed; month reported but age and year imputed and all information is
111
imputed are separately lower than 1 percent for all characteristics and in general for
all surveys. On the other hand, the responses where year reported but age and month
imputed is 1.3 %, 3.4 % and 1.5 % in general at TDHS-1993, TDHS-1998 and
TDHS-2003 respectively. The percent of such cases do not differ according to the
type of place of residence of women living. However, at East region, around 5 % and
3 % of the births for the last two surveys in that order, years are reported only
correctly and age and month is imputed. The highest percent for such situation at
TDHS-1993 is seen at North region with 2.1 %.
The percent distribution of children recorded at birth history section in terms of the
completeness of the birth date information is assessed by the characteristics of the
mother is presented at Table V.1.1.2. It is clearly seen that the complete information
is given by mothers aged 15-19. The women at the beginning of their reproductive
period give complete birth date data. As mostly, these births are the first of the
woman and at most 5 year will be passed after the birth of their children and this is a
short time to remember the exact birth date of their child. The completeness of the
information decreases when the age of the mother increases. It is better to keep up in
mind that information of the all live births are collected at birth history and a woman
at the 45-49 age group is asked to remember the birth date of her child regardless the
year the birth given. Sometimes woman was asked to give information about a birth
given 34 years ago. The fluctuation sourcing from TDHS-1998 results is seen among
all age groups except 15-19 where every birth is recorded completely. At the data
gathered from the others aged above 40, around 15 percent of the birth dates are
recorded as year and age reported but the month is imputed.
Table V.1.1.2 also presents the completeness percents of the date of birth information
of children at TDS-1993, TDHS-1998 and TDHS-2003 by the education of mother.
Results indicate that when the education level of mother increases, the completeness
of the birth information of the child also increases. While nearly complete data is
collected from the women who have secondary or more education (99.8 %, 99.3 %,
99.9 % respectively), the completeness is seen among the women have no education
or did not complete primary education. Dislike from other education groups, at the
112
group where the women completed secondary school or more, the completeness rate
did not decrease from TDHS-1993 to TDHS-2003.
113
Table V.1.1.1. Percent Distribution of Children Born by Completeness of
Information on Date of Birth by Region and Type of Place of Residence, TDHS
1993, 1998 and 2003
No
imputation
Month
and age
reported -
year
imputed
Year and
age
reported
- month
imputed
Year
reported
– age and
month
imputed
Age
reported
– year
and
month
imputed
Month
reported
– age and
year
imputed
All
imputed Total
Region
West
1993 97.6 0.0 1.2 0.9 0.0 0.0 0.3 5688
1998 90.5 0.0 5.3 3.3 0.6 0.0 0.3 5414
2003 96.6 0.0 2.1 0.9 0.2 0.0 0.1 7232
South
1993 97.0 0.0 2.0 0.9 0.0 0.0 0.1 3096
1998 89.5 0.0 8.2 1.9 0.2 0.0 0.1 2579
2003 91.3 0.9 4.9 1.5 1.1 0.1 0.2 2766
Central
1993 96.7 0.0 1.6 1.6 0.0 0.0 0.1 4668
1998 88.4 0.1 6.8 3.3 0.8 0.0 0.6 3895
2003 96.7 0.1 1.9 0.9 0.2 0.1 0.1 4691
North
1993 92.3 0.8 3.9 2.1 0.6 0.0 0.3 1851
1998 85.2 0.2 8.7 3.6 1.9 0.1 0.5 1500
2003 95.1 1.0 1.4 0.8 1.0 0.2 0.6 1583
East
1993 95.9 0.2 2.2 1.4 0.1 0.0 0.1 4524
1998 73.0 0.0 19.5 4.8 1.7 0.0 0.9 3821
2003 84.3 1.5 8.5 3.1 1.8 0.2 0.6 4901
Type of Place of Residence
Urban
1993 97.4 0.1 1.4 0.9 0.0 0.0 0.2 11332
1998 87.1 0.0 8.5 3.4 0.6 0.0 0.3 10644
2003 94.9 0.3 2.8 1.1 0.6 0.1 0.2 13986
Rural
1993 95.1 0.2 2.5 1.8 0.2 0.0 0.2 8495
1998 82.9 0.1 11.2 3.5 1.6 0.0 0.8 6565
2003 89.2 1.2 5.8 2.2 1.1 0.1 0.4 7187
Total
1993 96.4 0.1 1.9 1.3 0.1 0.0 0.2 19827
1998 85.5 0.1 9.5 3.4 0.9 0.0 0.5 17209
2003 93.0 0.6 3.8 1.5 0.8 0.1 0.3 21173
114
Table V.1.1.2. Percent Distribution of Children Born by Completeness of
Information on Date of Birth by Demographic Characteristics of Women,
TDHS 1993, 1998 and 2003
No
imputation
Month
and age
reported
-year
imputed
Year
and age
reported
- month
imputed
Year
reported
– age
and
month
imputed
Age
reported –
year and
month
imputed
Month
reported
– age
and
year
imputed
All
imputed Total
Age
15-19
1993 100.0 0.0 0.0 0.0 0.0 0.0 0.0 177
1998 100.0 0.0 0.0 0.0 0.0 0.0 0.0 159
2003 100.0 0.0 0.0 0.0 0.0 0.0 0.0 140
20-24
1993 99.0 0.0 0.5 0.2 0.0 0.0 0.3 1387
1998 98.0 0.0 1.4 0.6 0.0 0.0 0.0 1190
2003 99.3 0.0 0.5 0.2 0.0 0.0 0.0 1249
25-29
1993 98.9 0.0 0.6 0.5 0.0 0.0 0.0 2705
1998 95.0 0.1 3.5 0.9 0.4 0.0 0.2 2384
2003 97.1 0.2 1.9 0.5 0.1 0.0 0.1 2846
30-34
1993 97.9 0.2 1.1 0.5 0.1 0.1 0.1 3966
1998 90.0 0.1 7.3 1.8 0.5 0.1 0.2 3110
2003 95.2 0.4 2.8 1.2 0.2 0.0 0.1 3879
35-39
1993 96.9 0.0 1.7 1.3 0.0 0.0 0.1 4176
1998 85.8 0.0 9.2 3.4 1.0 0.0 0.7 3537
2003 93.0 0.9 3.2 1.5 0.9 0.3 0.3 4339
40-44
1993 94.1 0.3 3.0 1.9 0.3 0.0 0.5 4070
1998 79.7 0.1 14.0 4.0 1.4 0.0 0.8 3501
2003 90.4 1.0 5.0 1.8 1.2 0.1 0.5 4719
45-49
1993 93.7 0.1 3.3 2.6 0.2 0.0 0.2 3347
1998 75.2 0.1 14.9 7.4 1.6 0.0 0.8 3327
2003 88.6 0.4 6.7 2.6 1.3 0.1 0.3 4001
115
Table V.1.1.2. Percent Distribution of Children Born by Completeness of
Information on Date of Birth by Demographic Characteristics of Women,
TDHS 1993, 1998 and 2003 (Continued)
No
imputation
Month
and age
reported -
year
imputed
Year and
age
reported
- month
imputed
Year
reported
– age and
month
imputed
Age
reported –
year and
month
imputed
Month
reported
– age and
year
imputed
All
imputed
Total
Education
No educ/Pri. Inc.
1993 94.3 0.2 3.0 1.9 0.2 0.0 0.3 10042
1998 72.7 0.1 17.9 6.3 2.0 0.0 1.0 7378
2003 83.9 1.4 8.7 3.1 1.9 0.2 0.7 7435
Primary
1993 98.4 0.0 0.8 0.7 0.0 0.0 0.0 8776
1998 94.5 0.1 3.6 1.4 0.2 0.0 0.2 8603
2003 97.5 0.1 1.4 0.7 0.1 0.0 0.1 11732
Secondary
1993 99.8 0.0 0.1 0.1 0.0 0.0 0.0 1009
1998 99.3 0.0 0.2 0.5 0.0 0.0 0.0 1227
2003 99.9 0.0 0.1 0.0 0.0 0.0 0.0 2005
Total
1993 96.4 0.1 1.9 1.3 0.1 0.0 0.2 19827
1998 85.5 0.1 9.5 3.4 0.9 0.0 0.5 17209
2003 93.0 0.6 3.8 1.5 0.8 0.1 0.3 21173
Table V.1.1.3 shows the percentage of children with complete information on year
and month of birth by number of years since birth of child by region and type of
place of residence for all three TDHSs. During the taining, all interviewers were
especially told to record complete birth and dead date information for children born
within 5 years before the survey. Hence, nearly complete information is collected for
the 5 years preceding the surveys. Overall results show that ignorable level of
incompleteness is seen at the births 1, 2 and 4 years preceding the survey at TDHS-
1998 and 2, 3 and 4 years preceding the survey at TDHS-2003. While the
incompleteness is 0.1 % at TDHS-2003 for these years, 0.3, 0.2 and 0.4 percent of
the total birth date responses are calculated as incomplete at TDHS-1998 for the
years 1.2 and 4. For all three years as the birth date of the child is far from the
survey, the completeness rates are decreasing. Lowest completeness rates are seen at
116
the births given 20 or more years ago. The remembrance of the birth dates will go
down when the year is increasing. For the births 5 or more year ago, the
completeness of the month and year information decreases noticeable at North in
TDHS-1993 and East in TDHS-1998 and TDHS-2003. Fluctuation caused by 1998
TDHS results are also seen for all regions. On the other hand, more complete
information is seen at urban areas at the births happen 5 years or before the surveys.
Table V.1.1.4. indicates the completeness of the birth date data according to age and
education of the mother. For all births, mentioned at age 15-19, birth month and year
information is completely reported. The births occurred 4 or less year from the
survey is close to complete for all age groups of woman. However, the lowest
percent is seen at TDHS-1998 for the children born 4 or fewer years ago from the
survey with 88.5 %. The completeness for births happened 5 or more years ago
decreases especially after age 30. The highest completeness rates are seen among the
women aged 15 to 29. Educational difference is seen clearly at the completeness of
the birth date information with controlling the birth year of the children. It is seen
that for some surveys, except from the first 5 years, for some years more than 4 the
complete information is given by the women completed secondary education or
more. For all the births of these women, completeness of the information are high
than 98.0 ignoring the birth year of the children. Women having secondary or more
education gives 100.0 % and 99.6 % completeness rate at first two and at the last
respectively for the children born 20 or more year ago.
117
Table V.1.1.3. Percentage of Children with Complete Information on Year and
Month of Birth by Number of Years since Birth of Child by Region and Type of
Place of Residence, TDHS 1993, 1998 and 2003
Number of Years since Birth of Child
0 1 2 3 4 5-9 10-14 15-19 20+ Total
Region
West
1993 100.0 100.0 100.0 100.0 100.0 98.8 97.4 96.3 95.9 5688
1998 100.0 99.1 100.0 100.0 100.0 95.4 91.1 85.9 82.2 5414
2003 100.0 100.0 100.0 100.0 100.0 98.3 96.9 96.0 92.8 7232
South
1993 100.0 100.0 100.0 100.0 100.0 98.3 96.8 95.7 93.9 3096
1998 100.0 100.0 100.0 100.0 98.6 93.4 88.9 87.0 79.8 2579
2003 100.0 100.0 100.0 100.0 99.1 93.9 90.3 88.6 83.6 2766
Central
1993 100.0 100.0 100.0 100.0 100.0 98.1 96.8 95.7 92.8 4668
1998 100.0 100.0 100.0 100.0 100.0 92.4 85.6 85.5 78.6 3895
2003 100.0 100.0 100.0 100.0 100.0 99.1 96.6 95.3 93.9 4691
North
1993 100.0 100.0 100.0 100.0 100.0 94.1 92.5 88.3 86.0 1851
1998 100.0 100.0 100.0 100.0 99.3 92.2 85.5 78.7 73.4 1500
2003 100.0 100.0 100.0 100.0 100.0 95.7 93.5 94.5 93.2 1583
East
1993 100.0 100.0 100.0 100.0 100.0 96.7 95.4 93.8 91.0 4524
1998 100.0 100.0 99.3 100.0 99.3 74.3 64.5 62.6 55.0 3821
2003 100.0 100.0 99.7 99.8 100.0 87.0 77.5 77.7 72.6 4901
Type of Place
of Residence
Urban
1993 100.0 100.0 100.0 100.0 100.0 98.5 97.2 96.6 94.1 11332
1998 100.0 99.5 100.0 100.0 99.9 90.6 85.2 82.9 77.1 10644
2003 100.0 100.0 99.9 100.0 99.8 96.9 94.2 93.3 90.6 13986
Rural
1993 100.0 100.0 100.0 100.0 100.0 96.5 95.0 92.6 91.7 8495
1998 100.0 100.0 99.5 100.0 99.0 85.6 79.4 77.0 72.9 6565
2003 100.0 100.0 100.0 99.8 100.0 90.4 84.7 86.4 84.9 7187
Total
1993 100.0 100.0 100.0 100.0 100.0 97.6 96.2 94.8 93.0 19827
1998 100.0 99.7 99.8 100.0 99.6 88.7 82.9 80.7 75.5 17209
2003 100.0 100.0 99.9 99.9 99.9 94.7 90.9 90.9 88.7 21173
118
Table V.1.1.4. Percentage of Children with Complete Information on Year and
Month of Birth by Number of Years since Birth of Child by Demographic
Characteristics of Women, TDHS 1993, 1998 and 2003
Number of Years since Birth of Child
0 1 2 3 4 5-9 10-14 15-19 20+ Total
Age
15-19
1993 100.0 100.0 100.0 100.0 100.0 - - - - 177
1998 100.0 100.0 100.0 100.0 100.0 - - - - 159
2003 100.0 100.0 100.0 100.0 100.0 100.0 - - - 140
20-24
1993 100.0 100.0 100.0 100.0 100.0 95.5 49.3 - - 1387
1998 100.0 100.0 100.0 100.0 100.0 89.3 39.3 - - 1190
2003 100.0 100.0 100.0 100.0 100.0 95.8 46.3 - - 1249
25-29
1993 100.0 100.0 100.0 100.0 100.0 98.2 97.3 100.0 - 2705
1998 100.0 100.0 100.0 100.0 100.0 91.4 85.4 82.1 - 2384
2003 100.0 100.0 100.0 100.0 99.7 95.8 88.6 100.0 - 2846
30-34
1993 100.0 100.0 100.0 100.0 100.0 98.1 97.7 92.6 100.0 3966
1998 100.0 98.5 100.0 100.0 99.0 90.9 85.1 79.1 81.6 3110
2003 100.0 100.0 100.0 100.0 100.0 94.8 93.0 90.0 100.0 3879
35-39
1993 100.0 100.0 100.0 100.0 100.0 97.7 96.5 96.2 95.4 4176
1998 100.0 100.0 100.0 100.0 100.0 89.4 85.9 84.2 64.1 3537
2003 100.0 100.0 99.3 100.0 100.0 94.8 92.0 92.6 83.5 4339
40-44
1993 100.0 100.0 100.0 100.0 100.0 95.7 94.5 94.5 92.4 4070
1998 100.0 100.0 93.9 100.0 100.0 77.4 77.8 79.9 79.7 3501
2003 100.0 100.0 100.0 98.6 100.0 92.8 88.8 90.9 89.4 4719
45-49
1993 - 100.0 100.0 100.0 100.0 97.7 94.2 93.7 93.1 3347
1998 100.0 100.0 - 100.0 88.5 72.9 76.3 76.9 74.3 3327
2003 - 100.0 100.0 100.0 100.0 87.4 86.4 88.5 88.9 4001
119
Table V.1.1.4. Percentage of Children with Complete Information on Year and
Month of Birth by Number of Years since Birth of Child by Demographic
Characteristics of Women, TDHS 1993, 1998 and 2003 (Continued)
Number of Years since Birth of Child
0 1 2 3 4 5-9 10-14 15-19 20+ Total
Education
No educ/Pri Inc.
1993 100.0 100.0 100.0 100.0 100.0 96.0 94.6 92.7 90.6 10042
1998 100.0 98.8 99.3 100.0 99.4 75.6 68.6 68.6 65.1 7378
2003 100.0 100.0 99.7 99.8 99.6 85.7 79.9 80.7 79.5 7435
Primary
1993 100.0 100.0 100.0 100.0 100.0 98.8 97.8 97.5 97.4 8776
1998 100.0 100.0 100.0 100.0 99.6 95.7 93.5 91.7 89.0 8603
2003 100.0 100.0 100.0 100.0 100.0 98.5 96.8 96.6 95.4 11732
Secondary
1993 100.0 100.0 100.0 100.0 100.0 99.6 100.0 99.0 100.0 1009
1998 100.0 100.0 100.0 100.0 100.0 98.0 99.7 98.7 100.0 1227
2003 100.0 100.0 100.0 100.0 100.0 100.0 99.8 99.8 99.6 2005
Total
1993 100.0 100.0 100.0 100.0 100.0 97.6 96.2 94.8 93.0 19827
1998 100.0 99.7 99.8 100.0 99.6 88.7 82.9 80.7 75.5 17209
2003 100.0 100.0 99.9 99.9 99.9 94.7 90.9 90.9 88.7 21173
120
The sex of the child and the survival status of the children will have an effect on the
completeness of the birth date information of the children. Mothers may have in a
psychological situation by which they will forget the exact birth date of the dead
children. TDHS datasets are assessed for the completeness of the birth date data by
the survival status of the child and sex of the child with controlling the number of
years since the birth of child. Results are presented at Table V.1.1.5. It is clearly seen
that ignoring the births since 4 and less years before the surveys, the completeness of
the information on the birth date of the dead children is less complete than the living
ones. For all three surveys mother give more complete birth dates for the living
children as compared to the dead ones. On the other hand, female children’s birth
date information is more complete than the male children. With some exceptions, at
all three surveys, mothers remember their daughters’ birth dates better than their
sons.
Table V.1.1.5 also brings the estimation of completeness of the children’s birth dates
according to the time period of interviewer at the field with controlling the number of
years passed since the birth of child. The results show that there is no obvious
relationship between the time passed at the field and the completeness of the birth
date information. The interviewers focusing on the births of last 5 years and the
experience of them did not make any change on the quality of the date of birth data
of the children.
121
Table V.1.1.5. Percentage of Children with Complete Information on Year and
Month of Birth by Number of Years since Birth of Child by Survival Status and
Sex of Child and Time Period of Interviewer at the Field, TDHS 1993, 1998 and
2003
Number of Years since Birth of Child
0 1 2 3 4 5-9 10-14 15-19 20+ Total
Survival status of child
Dead
1993 100.0 100.0 100.0 100.0 100.0 91.8 88.9 85.7 84.1 2418
1998 100.0 92.8 100.0 100.0 100.0 68.9 62.1 56.8 54.7 1753
2003 100.0 100.0 97.5 100.0 96.8 80.2 70.3 75.8 75.3 1697
Alive
1993 100.0 100.0 100.0 100.0 100.0 98.2 97.2 96.4 95.4 17409
1998 100.0 100.0 99.8 100.0 99.5 90.1 84.9 84.1 80.2 15456
2003 100.0 100.0 100.0 99.9 100.0 95.5 92.5 92.3 91.1 19476
Sex of child
Male
1993 100.0 100.0 100.0 100.0 100.0 97.4 96.1 94.4 93.2 10219
1998 100.0 100.0 99.7 100.0 99.7 89.4 82.8 78.0 74.6 8837
2003 100.0 100.0 99.9 99.9 99.8 94.7 90.0 90.0 89.1 10818
Female
1993 100.0 100.0 100.0 100.0 100.0 97.8 96.4 95.2 92.8 9608
1998 100.0 99.4 100.0 100.0 99.5 88.0 83.0 83.5 76.5 8371
2003 100.0 100.0 100.0 100.0 100.0 94.7 91.8 91.8 88.3 10355
122
Table V.1.1.5. Percentage of Children with Complete Information on Year and
Month of Birth by Number of Years since Birth of Child by Survival Status and
Sex of Child and Time Period of Interviewer at the Field, TDHS 1993, 1998 and
2003 (Continued)
Number of Years since Birth of Child
0 1 2 3 4 5-9 10-14 15-19 20+ Total
Time period of
interviewer in
the field
1st week
1993 100.0 100.0 100.0 100.0 100.0 97.4 95.5 94.1 91.2 2504
1998 100.0 100.0 100.0 100.0 100.0 86.7 82.3 79.1 72.5 1945
2003 100.0 100.0 100.0 100.0 100.0 97.9 94.2 96.7 93.9 2613
2nd week
1993 100.0 100.0 100.0 100.0 100.0 98.2 96.7 94.1 88.8 2590
1998 100.0 100.0 100.0 100.0 100.0 90.8 86.1 78.6 71.2 1965
2003 100.0 100.0 100.0 99.4 100.0 96.3 94.8 92.5 90.8 2269
3rd week
1993 100.0 100.0 100.0 100.0 100.0 97.8 94.7 93.7 94.3 2996
1998 100.0 100.0 100.0 100.0 98.4 88.3 85.3 82.1 73.7 2332
2003 100.0 100.0 100.0 100.0 98.9 93.9 92.2 89.9 89.1 2448
4th week
1993 100.0 100.0 100.0 100.0 100.0 97.1 97.6 95.5 92.8 2518
1998 100.0 100.0 98.8 100.0 100.0 88.1 81.0 82.2 71.3 2348
2003 100.0 100.0 100.0 100.0 100.0 97.2 93.7 93.0 88.5 2526
More
1993 100.0 100.0 100.0 100.0 100.0 97.6 96.4 95.3 94.3 9218
1998 100.0 99.4 100.0 100.0 99.6 88.9 82.2 80.8 78.4 8620
2003 100.0 100.0 99.9 100.0 100.0 93.4 88.4 88.9 86.9 11318
Total
1993 100.0 100.0 100.0 100.0 100.0 97.6 96.2 94.8 93.0 19827
1998 100.0 99.7 99.8 100.0 99.6 88.7 82.9 80.7 75.5 17209
2003 100.0 100.0 99.9 99.9 99.9 94.7 90.9 90.9 88.7 21173
123
V.1.2. The Displacement of Children’s Birth Dates
The displacement of the children’s birth dates has also studied at the quality of the
birth history data. To escape from the workload of asking additional questions
especially at the section 4 and 5 of the questionnaire, interviewers may carry the birth
dates of the children out of the five year period. If there is a child written born inside
the 5 years preceding the survey, many questions had to be asked about the
pregnancy, delivery and early ages and current life of the children. Interviewers may
avoid asking these questions just change the birth date and carry child to the age 5 or
more. If the total fertility rate is estimated for 5 years preceding the survey, such kind
of a displacement of children will create a bias on the rate. Currently at most of the
countries, TFR at TDHS is estimated for 3 years period preceding the survey.
Therefore the effect of the displacement on TFR is not taken into consider at the
section: The Impact of Data Quality on Demographic Rates.
To understand “the heavy workload” of the interviewers, the number of questions at
Section 4 and 5 are shown at Table V.1.2.1. Questions about the last birth and the
previous births are evaluated in terms of whether they are directly asked to the
respondent or filled by the interviewer without asking. It is clearly seen that,
questions asked to the respondents are relatively high at TDHS-1998. While 95 and
85 questions were asked at TDHS-1993 and TDHS-2003 respectively, 127 questions
were placed to ask for the last birth. Although one out of three less questions was
asked for the previous birth, at the last two surveys, nearly same number of questions
was asked to last and previous births. On the other hand, the information filled in by
the interviewer is decreasing among the surveys. While 41 and 30 questions were
asked for the last and the previous births respectively, the number of questions
dropped to 22 for the last birth and 21 for the previous birth at TDHS-2003.
The median minutes to complete the ever-married questionnaire according to the
number of children under 5 is shown at Table V.1.2.2.. As expected, median time to
complete the questionnaire increases with the increase in numbers of the children
124
aged under 5. On the other hand, although the questions asked for the children under
5 increases the median time to complete the questionnaire decreases at TDHS-1998.
Table V.1.2.1. Number of Questions in Ever-Married Woman Questionnaires
that Depend on Children’s Year of Birth, TDHS 1993, 1998 and 2003
Questions about
Last Birth
Questions about
Previous Birth
Questions asked of respondents
1993 95 67
1998 127 121
2003 85 84
Other Information filled in by the interviewer
1993 41 30
1998 23 22
2003 22 21
Table V.1.2.2. Median Minutes to Complete the Ever-Married Woman
Questionnaire by the Number of Children Born in Last Five Years prior to
Survey, TDHS-2003.
Number of
Children
Median Minutes to Complete the Questionnaire
TDHS-1993 TDHS-1998 TDHS-2003
0 29 25 32
1 36 35 42
2+ 42 42 51
Total 33 30 36
125
The level of displacement of children is hard to calculate. The year of birth
distribution of children at all three surveys will help to identify the extent of
displacement If there is a significant displacement problem, the number of children
born 6 years prior to survey is more than 5 years. The birth year ratios for 4, 5 and 6
years prior to the survey shows the displacement of births from 5 to 6 years. In
theory, the birth ratios of 5 and 6 years must be around 100. If the ratio is over 100,
the number of children at this particular year is more than the year where the age
ratio is below 100.
Table V.1.2.3. indicates the number of births by calendar years and birth year ratios
for 4, 5, and 6 years prior to survey by region and type of place of residence the
respondent is living at the time of survey. For all surveys the number of children at
age 6 is more than 5. The highest displacement is seen at TDHS-1998. The gap
between the age ratio at 5 and 6 seems closed with the TDHS-2003. A vivid variation
is seen at the urban and rural areas for all three surveys. The displacement is high at
urban areas as compared with the rural. The highest gap between urban and rural is
seen at TDHS-1998. The gap between urban and rural in terms displacement closes
at TDHS-2003 as compared with the previous surveys. On the other hand, no clear
regional difference at level of displacement for all three surveys. The highest
displacement at TDHS-1993 is seen at East region; at TDHS-1998, West region has
the highest level of displacement and for the last TDHS, South region has the highest
displacement level.
Number of births by calendar years and birth year rations for 4, 5 and 6 years prior to
survey is calculated by age groups and the education of respondents whom the evermarried
women questionnaire is completed shown at Table V.1.2.4.. The least
difference is seen at the responses of the women aged 20-24 for the last two TDHSs
and women aged 25-29 at TDHS-1993. No clear relation is seen with the
displacement level and the age of respondent. The highest displacement is collected
from the women aged 45-49 at TDHS-1993 and TDHS-2003. At TDHS-1998 the age
group where the highest displacement level is seen among 40-44.
126
The lowest displacement is seen among the information collected from the primary
educated women for all surveys. It’s interesting that, women don’t have education or
did not completed primary education and the women having secondary or higher
education gave information where the displacement is seen high.
127
Table V.1.2.3. Number of Births by Calendar Years and Birth Year Ratios for
4, 5 and 6 Years prior to the Survey by Region and Type of Place of Residence,
TDHS 1993, 1998 and 2003
Years Prior to Survey
Birth Year Ratios centered
on Period prior to survey
0 1 2 3 4 5 6 7 8 9 4 yrs. 5 yrs. 6 yrs.
Region
West
1993 202 191 188 176 193 192 238 217 221 238 105.1 89.1 116.4
1998 209 229 189 188 215 185 253 186 226 190 115.5 79.0 136.3
2003 233 277 284 274 274 268 305 297 307 276 101.1 92.6 107.9
South
1993 110 110 102 119 116 134 119 118 125 131 91.2 114.5 94.2
1998 103 100 93 96 99 96 104 105 112 101 103.5 94.4 103.4
2003 118 94 110 117 119 100 121 105 138 117 109.3 83.4 118.1
Central
1993 153 176 133 174 147 148 195 191 178 173 91.0 86.9 114.8
1998 185 161 146 136 168 151 163 147 158 142 117.4 91.1 109.5
2003 134 142 183 174 179 179 186 166 160 180 101.4 98.3 107.4
North
1993 77 69 66 65 67 63 63 66 77 74 105.3 96.3 98.1
1998 56 54 53 53 55 58 54 52 64 50 99.9 105.8 98.2
2003 41 42 64 53 52 61 71 63 58 54 90.7 99.5 114.2
East
1993 206 169 163 182 174 155 254 229 217 180 103.2 72.5 132.2
1998 182 174 175 169 170 169 227 190 184 142 100.6 85.2 126.4
2003 233 220 227 246 242 211 245 246 227 219 105.9 86.9 106.9
Type of Place of Residence
Urban
1993 444 421 409 431 427 419 523 473 467 446 100.5 88.2 117.3
1998 453 439 415 406 448 398 517 412 476 382 111.4 82.6 127.6
2003 490 513 582 578 559 551 621 571 584 566 99.0 93.3 110.8
Rural
1993 305 295 244 286 270 273 346 348 351 350 96.5 88.9 111.1
1998 281 278 241 236 260 261 284 268 267 243 104.8 95.8 107.4
2003 268 263 286 286 307 269 306 308 305 280 110.6 87.8 106.2
Total
1993 749 716 653 717 697 692 869 821 817 795 98.9 88.5 114.8
1998 735 718 656 642 709 659 801 680 743 625 108.9 87.3 119.6
2003 759 776 868 864 866 820 927 878 890 846 102.8 91.4 109.2
128
Table V.1.2.4. Number of Births by Calendar Years and Birth Year Ratios for
4, 5 and 6 Years prior to the Survey by Demographic Characteristics of Women,
TDHS 1993, 1998 and 2003
Years Prior to Survey
Birth Year Ratios centered
on Period prior to survey
0 1 2 3 4 5 6 7 8 9 4 yrs. 5 yrs. 6 yrs.
Age
15-19
1993 87 40 27 5 2 - - - - - - - -
1998 78 45 22 8 4 1 1 - - - 87.1 20.5 -
2003 69 37 18 13 2 1 - - - - 31.2 - -
20-24
1993 271 258 194 196 145 85 87 34 16 7 103.5 72.9 146.5
1998 247 225 206 157 138 103 61 31 15 3 105.6 103.9 91.3
2003 279 233 229 179 139 97 50 26 12 4 100.9 102.0 82.1
25-29
1993 207 219 218 240 238 262 266 226 201 149 94.9 103.8 109.0
1998 206 226 223 241 256 223 250 202 205 118 110.3 88.3 117.5
2003 216 275 289 300 308 276 268 255 211 166 106.9 95.9 100.8
30-34
1993 125 121 129 167 191 205 266 300 280 302 102.6 89.6 105.5
1998 128 137 132 134 177 180 245 221 242 245 113.0 85.2 122.4
2003 126 137 205 235 237 220 312 294 317 311 104.1 80.0 121.7
35-39
1993 45 56 59 71 84 88 147 155 186 179 105.6 76.5 120.5
1998 63 63 51 72 92 100 145 123 183 141 107.4 83.9 130.7
2003 54 71 85 92 125 155 195 196 208 213 101.1 97.1 111.1
40-44
1993 15 18 21 34 30 47 76 82 93 114 73.4 89.4 117.6
1998 12 20 20 23 31 40 77 75 74 80 99.2 74.0 133.8
2003 15 20 34 40 41 62 82 80 109 123 80.5 100.3 115.1
45-49
1993 - 3 5 4 7 6 27 24 40 44 152.2 35.0 181.6
1998 1 2 - 7 11 13 21 29 23 38 110.2 78.4 102.2
2003 - 3 8 5 13 9 19 26 32 29 193.6 56.3 108.5
129
Table V.1.2.4. Number of Births by Calendar Years and Birth Year Ratios for
4, 5 and 6 Years prior to the Survey by Demographic Characteristics of Women,
TDHS 1993, 1998 and 2003 (Continued)
Years Prior to Survey
Birth Year Ratios centered
on Period prior to survey
0 1 2 3 4 5 6 7 8 9 4 yrs. 5 yrs. 6 yrs.
Education
No educ/Pri.Inc.
1993 247 248 219 275 277 268 376 372 361 331 102.0 82.1 117.5
1998 179 169 183 197 208 182 315 237 287 241 109.7 69.7 150.0
2003 189 186 227 239 258 224 292 270 290 269 111.2 81.6 118.3
Primary
1993 431 400 367 378 353 368 421 398 401 408 94.5 95.2 109.8
1998 474 453 392 378 420 415 408 381 406 328 105.9 100.3 102.4
2003 463 450 501 511 493 491 516 517 526 499 98.4 97.3 102.3
Secondary
1993 72 68 68 64 67 56 73 51 55 57 112.1 80.6 134.5
1998 81 96 81 67 80 62 79 62 50 55 125.1 77.6 127.7
2003 107 139 140 114 115 104 119 91 75 78 105.4 89.1 121.8
Total
1993 749 716 653 717 697 692 869 821 817 795 98.9 88.5 114.8
1998 735 718 656 642 709 659 801 680 743 625 108.9 87.3 119.6
2003 759 776 868 864 866 820 927 878 890 846 102.8 91.4 109.2
Table V.1.2.5. shows the number of births by calendar years and the birth year ratios
calculated for 4, 5 and 6 years prior to survey by the survival status and the sex of
child and time period of interviewer in the field. While no displacement of children
to 6 years prior to survey is seen among dead children at TDHS-1993 and TDHS-
1998, for the last survey results show that a clear displacement is seen among the
dead children. Although, among the children who are alive at the time of survey, the
displacement is seen for all surveys; the level of displacement decreased with the
results of TDHS-2003 when compared with the previous surveys.
130
The displacement of the births to 6 year prior to survey is seen all surveys regarding
the sex of children. The significantly high displacement is seen at the female children
at TDHS-1998. While the displacement level did not change among male children,
among female children the displacement decreased at TDHS-2003. The time period
for the interviewer at the field seems have relation with the displacement level of the
children born 5 years prior to survey to 6 years at TDHS-1993 and TDHS-1998.
While the time the interviewer stays at the field increases, the level of displacement
also increases. TDHS-2003 results are dissimilar than the previous surveys, the
highest displacement is seen among the data which the interviewer collected at their
first week at the field.
Table V.1.2.5. Number of Births by Calendar Years and Birth Year Ratios for
4, 5 and 6 Years prior to the Survey by Survival Status and Sex of Child and
Time Period of Interviewer in the field, TDHS 1993, 1998 and 2003
Years Prior to Survey
Birth Year Ratios centered
on Period prior to survey
0 1 2 3 4 5 6 7 8 9 4 yrs. 5 yrs. 6 yrs.
Survival status of child
Dead
1993 27 47 42 40 48 65 79 71 78 108 91.8 101.9 116.6
1998 23 28 44 30 35 43 51 46 43 45 94.2 101.2 113.3
2003 15 27 23 35 32 26 44 51 55 58 105.9 67.5 114.1
Alive
1993 749 716 653 717 697 692 869 821 817 795 98.9 88.5 114.8
1998 712 689 612 612 674 616 750 634 700 580 109.8 86.5 120.1
2003 743 749 845 829 833 794 884 827 834 788 102.7 92.5 109.0
Sex of child
Male
1993 381 385 322 363 366 347 439 434 402 431 103.1 86.2 112.4
1998 377 372 353 336 358 350 397 351 392 314 104.3 92.7 113.3
2003 396 392 460 461 423 416 477 425 444 439 96.5 92.4 113.4
Female
1993 368 331 331 354 331 346 430 387 415 365 94.7 90.9 117.4
1998 357 346 303 306 350 309 404 329 352 310 114.0 82.0 126.5
2003 363 383 408 403 442 404 450 453 445 407 109.6 90.4 105.1
131
Table V.1.2.5. Number of Births by Calendar Years and Birth Year Ratios for
4, 5 and 6 Years prior to the Survey by Survival Status and Sex of Child and
Time Period of Interviewer in the field, TDHS 1993, 1998 and 2003 (Continued)
Years Prior to Survey
Birth Year Ratios centered
on Period prior to survey
0 1 2 3 4 5 6 7 8 9 4 yrs. 5 yrs. 6 yrs.
Time period of interviewer in the field
1st week
1993 98 102 75 80 85 92 99 108 80 90 99.1 100.2 99.0
1998 95 82 68 78 75 75 81 85 79 85 97.7 96.1 101.2
2003 79 82 101 104 128 82 122 99 103 106 137.8 66.1 134.3
2nd week
1993 95 79 80 77 94 101 109 107 104 112 105.4 99.4 105.0
1998 90 83 81 69 88 73 85 81 84 64 124.1 84.0 110.8
2003 72 92 55 90 66 95 90 85 95 88 71.9 121.4 99.9
3rd week
1993 118 120 85 117 108 124 135 124 123 116 89.4 102.1 109.1
1998 89 96 104 99 101 109 118 95 99 75 96.5 99.9 115.7
2003 94 85 84 97 91 93 113 112 109 89 96.3 91.1 110.1
4th week
1993 92 84 98 89 96 74 115 101 112 105 117.5 70.3 131.5
1998 106 104 99 92 103 85 106 86 102 76 117.1 81.4 123.6
2003 94 91 119 86 99 88 96 97 95 114 113.6 90.4 103.6
More
1993 347 331 316 354 314 301 410 381 397 373 96.0 83.1 120.2
1998 354 353 304 304 342 317 411 333 380 326 110.1 84.2 126.4
2003 418 425 510 488 481 461 507 486 488 449 101.3 93.4 107.1
Total
1993 749 716 653 717 697 692 869 821 817 795 98.9 88.5 114.8
1998 735 718 656 642 709 659 801 680 743 625 108.9 87.3 119.6
2003 759 776 868 864 866 820 927 878 890 846 102.8 91.4 109.2
132
The percent of births by years prior to survey for 10 years is summarized with Figure
V.1.2.1.. The fluctuation at the distribution of births for the years prior to survey is
high at TDHS-1998. Especially the decrement from year 4 to 5 and increment from 5
to 6 is sharpest at TDHS-1998.The sharp decreases seen for year 7 and 9 at TDHS-
1998. The least fluctuation, on the other hand, is seen at TDHS-2003 data. The peak
at year 6 is lowest for this survey.
Figure V.1.2.1. Percent Distribution of Births for Ten Years prior to Survey,
TDHS 1993, 1998 and 2003
V.1.3. Age Heaping
Age heaping can be a common problem for the births mentioned by the mothers at
the birth history section of the DHS questionnaires especially for the undeveloped
countries. Like the age data at household list and at the beginning of the individual
questionnaire, heaping the ages to certain years will be a problem of birth history
data. The single age ratios will help to illustrate the problem of heaping. Table
133
V.1.3.1. shows the age ratios surviving children up to age 15 by region and type of
place of residence at TDHS-1993, 1998 and 2003. In general, a heaping to age 6 is
seen for all three surveys. The interviewers may carry the children aged 5 to age 6 to
escape from the workload of asking several questions at Section 4 and 5. On the
other hand at TDHS-1998, a heaping for age 8 is seen. It will be a problem of
heaping the birth year of the children to the year 1990 which results in a heaping on
age 8. Except from age 6, the highest single age ratios are seen for age 13 which will
be a result of mentioning the birth years of the children as 1980 and 1990
respectively at TDHS-1993 and TDHS-2003.
Heaping at age 6 is seen nearly at all regions. The highest heaping to age 6 is seen at
East region at TDHS-1993, at West region at TDHS-1998 and at South region at
TDHS-2003. While the heaping on age 6 is decreasing at East, Central and West
regions from TDHS1993 to TDHS-2003, for South and North, the heaping problems
seems increasing for age 6. In addition, at TDHS-1998, a clear heaping on age 8 is
seen especially at West and North region. At South region the heaping of age 8 and
13 is vivid at TDHS-2003. For the same survey, at Region North age heaping for age
2 is remarkable.
The heaping problem for age 6 is seen more in urban data than the rural. Ignoring the
TDHS-1998 urban data, the heaping seems decreasing. Besides the heaping at age 6,
at TDHS urban data, the heaping of age 4 and age 8 is remarkable. At TDHS-1993
and TDHS-2003 urban data the highest heaping is seen at age 13. At TDHS-2003
rural data, on the other hand, the highest heaping is seen at age 4. The highest
heaping at TDHS-1993 rural data is seen at age 13.
134
Table V.1.3.1. Age Ratios for Living Children by Single Year of Age by Region
and Type of Place of Residence, TDHS 1993, 1998 and 2003
Age Ratio Centered on Age
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Region
West
1993 97.8 102.7 92.2 105.1 89.1 116.4 94.6 97.0 105.2 100.3 95.5 99.2 112.3 94.6 21.4
1998 115.2 90.7 92.9 115.5 79.0 136.3 77.8 120.0 86.6 105.3 105.8 88.0 115.5 87.5 19.2
2003 107.4 102.9 98.2 101.1 92.6 107.9 97.2 107.1 96.6 88.0 122.1 86.4 111.2 102.6 14.0
South
1993 104.0 88.6 109.9 91.2 114.5 94.2 96.8 100.3 105.3 97.9 97.0 99.1 116.7 82.6 25.1
1998 101.8 95.1 99.8 103.5 94.4 103.4 97.4 108.8 89.6 110.7 89.8 110.8 103.6 89.7 18.5
2003 82.3 104.4 102.3 109.3 83.4 118.1 81.2 123.9 88.9 105.9 106.6 79.6 129.7 90.8 18.0
Central
1993 123.4 75.9 124.6 91.0 86.9 114.8 102.4 97.8 97.3 96.4 106.2 101.5 99.0 97.5 24.6
1998 97.2 98.2 86.5 117.4 91.1 109.5 91.5 109.2 97.4 95.1 92.5 118.0 87.3 118.2 16.5
2003 89.6 116.0 96.1 101.4 98.3 107.4 96.1 92.6 111.9 97.0 96.7 91.9 107.1 105.1 16.1
North
1993 96.6 99.1 96.8 105.3 96.3 98.1 94.8 109.6 91.7 116.0 88.6 96.6 125.8 78.6 26.1
1998 98.7 99.7 97.7 99.9 105.8 98.2 88.4 125.0 85.2 96.3 103.9 107.4 94.4 103.8 17.6
2003 80.7 132.9 92.5 90.7 99.5 114.2 98.1 98.6 99.0 97.1 87.7 115.2 93.3 108.2 18.6
East
1993 91.4 93.0 108.0 103.2 72.5 132.2 97.2 106.2 83.2 122.7 81.7 114.8 106.2 85.1 28.9
1998 97.6 101.9 98.1 100.6 85.2 126.4 92.5 110.8 81.3 112.3 92.3 112.0 89.0 115.4 19.7
2003 95.7 97.6 104.7 105.9 86.9 106.9 104.5 97.5 99.1 101.3 97.2 104.5 94.7 108.2 21.8
Type of Place of Residence
Urban
1993 98.7 96.0 103.1 100.5 88.2 117.3 95.5 101.6 94.4 105.8 95.2 102.7 107.3 94.5 23.6
1998 101.2 98.1 94.1 111.4 82.6 127.6 83.0 119.9 85.3 108.0 94.8 103.2 100.9 97.5 18.8
2003 95.6 106.7 101.4 99.0 93.3 110.8 94.7 102.8 100.1 97.7 107.1 85.2 115.7 103.0 15.5
Rural
1993 107.5 84.1 111.3 96.5 88.9 111.1 100.1 100.5 99.1 104.8 92.4 103.1 113.2 83.1 25.9
1998 106.5 93.8 94.1 104.8 95.8 107.4 97.3 104.6 92.3 99.9 99.9 107.5 95.7 108.1 17.7
2003 94.8 104.3 96.4 110.6 87.8 106.2 100.6 104.0 97.3 93.2 103.8 107.4 92.0 103.7 19.4
Total
1993 102.1 91.2 106.2 98.9 88.5 114.8 97.4 101.1 96.4 105.4 94.0 102.9 109.7 89.6 24.6
1998 103.2 96.5 94.1 108.9 87.3 119.6 88.1 113.9 87.9 104.7 96.8 104.9 98.7 101.5 18.4
2003 95.4 105.9 99.7 102.8 91.4 109.2 96.7 103.2 99.1 96.1 105.9 92.6 107.0 103.2 16.8
Table V.1.3.2. presents the age ratios of living children by single year of age by age,
and education of the women. When the heaping is assessed by the age of women, it
is hard to mention a relationship between the age of woman and the heaping of
certain digits. The digits like 6, 4, and 8 and 13 are preferred by nearly all women
with different magnitudes. Heaping of digit 6 highly effected by the displacement of
the births from age 5 to age 6. Therefore this heaping has to be evaluated with
keeping the displacement factor in mind. When the heaping s evaluated with the
education of the woman respondent, it is seen that the relationship with the heaping
problem and the level of education of women is not vivid. Different digits were
135
preferred by women with different level education lowest heaping on certain digit 6
is seen among women having primary education only. The heaping on digit 8 is
highest at TDHS-1998 at women did not have education or did not complete primary
education. In addition, the heaping on digit 13 is seen as highest at TDHS-2003
among women having secondary or more education.
The age ratios of living children of single age by the sex of children and the time
period of interviewer in the field is shown at Table V.1.3.3.. The heaping at age 8 is
not seen at the data collected at the interviewers’ first week at the field. The age ratio
on age reaches to highest levels at 2nd week at TDHS-2003, and 4th week at TDHS-
1993 and TDHS-1998. Like age 8, the heaping on age 13 is not seen at the data
collected at the first week of the interviewer at the field. The heaping on age 13 is
seen at 3rd week for all three surveys. The sex of the child, on the other hand, does
not have direct effect on the heaping of the certain ages like 6, 8 and 13. Heaping on
age 8 is seen both males and females at TDHS-1998. While no heaping on this age is
seen at TDHS-2003, only for female children at TDHS-1993, heaping is vivid.
Heaping on age 13 is seen at female children’s ages at TDHS-1993 and TDHS-2003
only. Results show that the heaping on age 13 decreases among male children
gradually among the surveys.
136
Table V.1.3.2. Age Ratios for Living Children by Single Year of Age by
Demographic Characteristics of Women, TDHS 1993, 1998 and 2003
Age Ratio Centered on Age
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Age
15-19
1993 70.7 118.0 36.8 - - - - - - - - - - - -
1998 89.6 84.3 59.7 87.1 20.5 - - - - - - - - - -
2003 85.8 69.9 135.3 31.2 - - - - - - - - - - -
20-24
1993 111.1 85.4 115.5 103.5 72.9 146.5 66.4 78.5 77.3 - - - - - -
1998 99.4 107.8 91.5 105.6 103.9 91.3 80.8 89.3 38.3 - - - - - -
2003 91.8 111.2 97.1 100.9 102.0 82.1 84.1 78.3 54.0 - - - - - -
25-29
1993 103.2 94.9 105.3 94.9 103.8 109.0 96.8 107.4 91.6 111.1 84.8 110.3 62.9 48.8 149.3
1998 105.1 95.8 100.4 110.3 88.3 117.5 88.8 128.4 77.5 109.6 90.9 92.7 86.1 64.9 89.6
2003 108.8 100.6 100.5 106.9 95.9 100.8 106.6 100.2 102.5 92.5 95.0 95.5 94.4 - -
30-34
1993 95.4 89.8 104.3 102.6 89.6 105.5 109.7 93.0 105.5 100.2 98.3 107.1 109.9 92.5 74.8
1998 104.9 97.9 86.5 113.0 85.2 122.4 90.7 103.8 95.4 114.0 97.1 105.8 88.8 113.4 69.7
2003 82.9 110.1 106.5 104.1 80.0 121.7 93.4 104.8 100.2 99.1 107.2 100.1 100.3 106.2 75.5
35-39
1993 109.0 92.8 98.9 105.6 76.5 120.5 93.4 111.2 86.0 114.8 88.5 104.7 114.6 84.6 43.8
1998 110.4 76.0 100.5 107.4 83.9 130.7 74.7 138.8 79.3 98.2 105.5 98.3 107.8 95.4 35.6
2003 101.3 105.0 87.1 101.1 97.1 111.1 97.3 101.7 99.6 94.7 102.2 95.1 108.5 106.6 34.6
40-44
1993 98.9 80.5 134.6 73.4 89.4 117.6 96.6 95.3 107.1 96.2 95.9 94.6 114.4 95.0 20.3
1998 123.4 94.4 90.4 99.2 74.0 133.8 98.7 96.5 93.1 97.5 93.4 108.3 104.8 104.4 13.2
2003 82.3 113.1 107.2 80.5 100.3 115.1 84.1 107.6 101.2 95.2 107.1 88.8 113.6 90.1 14.9
45-49
1993 - 151.1 60.0 152.2 35.0 181.6 70.1 119.9 84.2 110.5 105.5 95.6 99.0 92.9 10.0
1998 - - - 110.2 78.4 102.2 131.2 69.6 125.5 86.2 81.7 126.4 86.9 101.1 8.6
2003 - 210.5 44.8 193.6 56.3 108.5 101.5 117.0 75.5 95.5 134.0 65.7 109.6 123.2 5.6
Education
No educ/Pri. Inc.
1993 106.6 83.6 111.0 102.0 82.1 117.5 100.9 102.8 87.8 110.1 93.5 102.7 112.1 88.4 22.8
1998 93.1 100.3 100.7 109.7 69.7 150.0 78.9 120.0 86.6 101.1 101.2 103.8 94.9 111.4 15.4
2003 89.5 107.0 98.6 111.2 81.6 118.3 92.7 107.6 90.9 107.2 99.9 93.1 107.1 105.8 15.6
Primary
1993 100.2 94.3 105.1 94.5 95.2 109.8 96.9 99.5 101.5 104.0 93.6 103.6 107.0 91.8 26.7
1998 104.6 94.3 93.2 105.9 100.3 102.4 93.7 114.5 85.6 109.7 92.1 107.6 102.0 91.5 21.6
2003 93.4 104.2 102.9 98.4 97.3 102.3 99.4 103.4 104.4 86.7 113.1 93.0 104.3 101.7 17.6
Secondary +
1993 98.1 102.5 94.5 112.1 80.6 134.5 80.6 101.8 122.1 81.2 103.4 97.3 107.8 83.5 31.5
1998 118.5 99.1 82.9 125.1 77.6 127.7 95.7 85.5 113.1 92.0 106.5 93.0 101.7 109.3 26.1
2003 112.8 110.5 89.2 105.4 89.1 121.8 94.4 87.9 97.9 117.8 87.7 88.3 124.2 101.5 18.1
Total
1993 102.1 91.2 106.2 98.9 88.5 114.8 97.4 101.1 96.4 105.4 94.0 102.9 109.7 89.6 24.6
1998 103.2 96.5 94.1 108.9 87.3 119.6 88.1 113.9 87.9 104.7 96.8 104.9 98.7 101.5 18.4
2003 95.4 105.9 99.7 102.8 91.4 109.2 96.7 103.2 99.1 96.1 105.9 92.6 107.0 103.2 16.8
137
Table V.1.3.3. Age Ratios for Living Children by Single Year of Age by Sex of
the Child and Time Period of Interviewer in the Field, TDHS 1993, 1998 and
2003.
Age Ratio Centered on Age
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Sex of Child
Male
1993 109.5 86.2 105.5 103.1 86.2 112.4 103.2 93.0 103.7 100.8 98.1 101.2 108.6 88.3 25.3
1998 101.8 99.8 94.6 104.3 92.7 113.3 89.0 117.7 83.7 108.1 97.7 101.6 105.1 92.6 19.0
2003 91.7 107.8 104.4 96.5 92.4 113.4 92.3 102.9 102.7 96.3 98.5 101.0 103.6 101.9 17.0
Female
1993 94.8 96.6 106.9 94.7 90.9 117.4 91.6 110.4 89.0 110.7 89.7 104.6 110.8 91.0 23.8
1998 104.8 92.9 93.6 114.0 82.0 126.5 87.1 110.0 92.6 101.2 95.9 108.5 92.5 111.4 17.9
2003 99.5 103.8 94.8 109.6 90.4 105.1 101.2 103.6 95.5 96.0 113.9 84.4 110.7 104.7 16.6
Time period of interviewer in the field
1st week
1993 117.7 82.9 99.2 99.1 100.2 99.0 120.5 81.1 100.8 101.8 101.6 106.3 94.0 101.6 23.8
1998 100.6 84.8 109.7 97.7 96.1 101.2 105.9 93.6 104.2 101.0 95.0 106.4 95.6 105.7 20.6
2003 91.5 108.1 90.6 137.8 66.1 134.3 87.8 100.6 97.2 108.7 97.4 95.6 110.9 93.9 16.4
2nd week
1993 90.2 102.9 88.6 105.4 99.4 105.0 100.6 95.1 110.4 90.8 99.9 105.1 110.3 86.9 22.3
1998 96.9 106.7 81.7 124.1 84.0 110.8 96.1 115.5 86.4 85.8 116.4 109.6 74.0 130.4 19.9
2003 144.5 60.4 148.2 71.9 121.4 99.9 92.0 109.6 93.6 95.7 128.7 73.8 125.7 87.5 14.4
3rd week
1993 119.2 71.1 121.9 89.4 102.1 109.1 95.7 102.8 93.0 110.0 89.8 101.0 116.3 91.4 22.9
1998 99.1 106.9 96.9 96.5 99.9 115.7 87.8 116.0 80.8 98.6 120.1 81.2 118.8 94.9 18.3
2003 95.4 92.7 110.2 96.3 91.1 110.1 100.9 108.2 92.0 87.5 118.7 85.0 121.0 89.8 19.3
4th week
1993 88.4 113.2 91.8 117.5 70.3 131.5 88.8 109.1 91.6 118.9 82.4 106.3 112.1 85.1 25.8
1998 101.5 101.2 90.5 117.1 81.4 123.6 82.8 125.5 79.4 107.7 96.4 103.7 108.0 85.5 22.1
2003 85.3 134.3 79.1 113.6 90.4 103.6 101.5 90.3 120.3 86.0 114.0 79.6 125.8 99.9 16.5
More
1993 100.0 92.1 112.4 96.0 83.1 120.2 94.3 105.4 94.3 105.6 95.1 101.1 111.0 88.0 25.6
1998 107.2 92.5 94.1 110.1 84.2 126.4 84.2 115.4 88.5 111.0 88.0 111.1 98.4 100.9 16.7
2003 91.7 111.6 98.5 101.3 93.4 107.1 97.7 104.4 97.8 97.7 99.2 101.6 96.2 113.0 17.0
Total
1993 102.1 91.2 106.2 98.9 88.5 114.8 97.4 101.1 96.4 105.4 94.0 102.9 109.7 89.6 24.6
1998 103.2 96.5 94.1 108.9 87.3 119.6 88.1 113.9 87.9 104.7 96.8 104.9 98.7 101.5 18.4
2003 95.4 105.9 99.7 102.8 91.4 109.2 96.7 103.2 99.1 96.1 105.9 92.6 107.0 103.2 16.8
138
V.1.4. Miscalculation of Year of Birth
During the field study, either by the respondent or by the interviewer the year of birth
of the child may be miscalculated. When the age of the child is known but the year
and month of the child is unknown, interviewer may just subtract the age from the
year of interview and calculate the year of birth. This calculation is true if the month
of birth is earlier than the month of interview. If the month of interview is earlier
than the month of birth, the year of birth will be overestimated for one year. If the
imputation is done at the field it is hard to understand , however if the imputation is
done at data entry process it will be more clear. An uneven distribution of birth will
show extent of this problem. At this section of the study the month of interview is
used as a basis to understand the level of problem.
The percent of the child whose month of birth falls in the month of interview or
earlier is shown on Table V.1.4.1.. The assessment is done for imputed and nonimputed
cases separately to see the effect of imputation during the data entry. The
results indicate that, the imputed cases at TDHS-1998 are around 5.5 times more
from TDHS-1993 and nearly 2 times than TDHS-2003. No clear distortion at the
distribution of births is seen at TDHS-2003. At TDHS-1993 and TDHS-1998, the
percent of the expected births earlier than the interview date is less than the actual
cases. There is around 15 % difference between the actual and expected percents
among imputed cases. This is only around 6 percent for non-imputed cases. The
imputation of the birth months in general is noticeably high which will result in a
problem of calculating the year of birth of children to an earlier year. The
comparison of the non-imputed cases to the imputed cases shows that the distortion
seems a problem of imputed cases.
The regional difference is seen at the problem of miscalculation of year of birth. At
TDHS-1993 data, the lowest gap between the expected and actual percentage is seen
at West region. The highest problem is seen at South region at imputed cases for the
same survey. North region seems the biggest problematic region a TDHS-1998 with
nearly 20 % difference between the expected and actual distribution of the birth
139
months. TDHS-2003 birth history data on birth date of children seems problematic
for Central and East regions where the expected months are higher than the actual
months given by the imputation process.
The data collected at urban areas needed to be imputed more than the data of rural
areas at TDHS-1998. For the last and the first TDHS the imputed cases are high in
numbers at rural areas than urban. No clear difference is seen at the expected and
actual birth month distribution of the births for imputed cases.
Same assessments were done by age and education of the respondent and time period
of the interviewer at the field. Table VIII.3.1. and Table VIII.3.2. at Annex VIII.3
indicates the calculation of percentage of children whose month of birth falls in the
month of interview or earlier by the age and education of mother and the time of the
interviewer at the field. As there is no direct relationship between the data collected
from the different age groups of women, their education and the time period of the
interviewer at the field and the level of distortion at the birth date data imputed at the
data entry process, the tables will only be used to evaluated for the evaluation of
number of cases imputed and their change with the selected demographic
characteristics of women and the interviewer.
140
Table V.1.4.1. Percentage of Children Whose Month of Birth Falls in the Month
of Interview or Earlier by Region and Type of Place of Residence, TDHS 1993,
1998 and 2003
Imputed Cases Number of
Children
Non-Imputed Cases Number of
Actual Expected Actual Expected Children
Region
West
1993 75.0 70.7 134 77,2 72.9 5554
1998 84.3 70.4 514 74.7 71.4 4901
2003 65.1 67.3 243 72.0 71.5 6989
South
1993 92.4 75.8 91 79.0 73.8 3005
1998 88.2 74.4 270 80.2 76.7 2308
2003 82.5 85.6 214 84.6 83.8 2552
Central
1993 91.3 75.8 155 82.7 75.7 4513
1998 89.9 76.0 447 78.7 75.1 3447
2003 30.7 40.1 148 45.1 44.4 4544
North
1993 88.5 73.1 127 73.4 72.1 1723
1998 90.4 71.9 219 75.4 71.3 1280
2003 15.1 17.6 59 16.3 16.7 1524
East
1993 85.4 70.5 176 80.9 72.1 4348
1998 88.6 72.3 1030 77.3 69.9 2792
2003 24.9 34.5 686 37.1 35.6 4215
Type of Place of Residence
Urban
1993 83.1 72.1 287 78.4 73.2 11045
1998 88.7 73.3 1365 76.9 73.3 9279
2003 45.4 51.2 667 58.7 58.1 13319
Rural
1993 88.5 73.6 396 80.5 73.8 8099
1998 87.3 72.1 1115 77.3 71.9 5450
2003 37.6 45.6 682 49.8 48.6 6505
Total
1993 86.2 72.9 683 79.3 73.5 19144
1998 88.1 72.8 2480 77.1 72.8 14729
2003 41.5 48.4 1349 55.8 55.0 19824
141
V.1.5. Coverage of Live Births
The DHS program improved the fertility section of the survey with previous
experiences. The latest reproduction section starts with asking the number of children
born, number of children living with the mother, number of children away separately
for both sexes. With brass type questions the basic aim is not to miss any of the
children born whether live or dead. To ask the question for each sex strengthens the
quality of data in terms of not missing any children at the birth history. After the
birth history completed, the interviewer sums up all the births mentioned at the birth
history section and a follow up question is asked to ascertain the number of live
births of the woman.
The underreporting of the dead children is one of the common problems in
developing countries especially the child was dead in neo-natal period. The omission
of births are hard to catch is there is no huge number of underreporting To calculate
the underreporting in a birth history data average number of children ever born by
age of mother in groups can be used. In a situation where no gross underreporting is
seen, the average parity should increase monotonically with the age groups.
Table V.1.5.1. indicates the average number of children ever born by age of mother
by region and type of place of residence for all three surveys. Results indicate that,
monotonical increase in average number of births is at least “0.4” among all age
groups at all surveys except a “0.2” increase at TDHS-2003 for the last two age
groups. Coverage of living children problem may be at this survey at the age group
45-49.. In addition, at the West region for the same survey the increase at parity
among the ages 30-34 and 35-39 and for the last two age groups is calculated as 0.3
and 0.1 respectively. Again at TDHS-2003, birth history data collected at North
region brings into matter a coverage problem for 305-39 and 40-44; and 40-44 and
45-49 age groups with an increase of 0.3and 0.2 in that order. The only minus value
is calculated for all surveys is at North region at TDHS-1993 for the last two age
groups. For the last two eligible age groups, the average number of children is below
142
0.4 at urban areas at TDHS-1993 and TDHS-2003 and at rural areas at TDHS-1998
and TDHS-2003.
The median age at first birth is also used to identify the coverage of live births. If the
median age of first birth did not changed significantly over time the age at first birth
of the cohorts will be more or less same. The assessment of median age at first birth
by age of woman at the time of survey by region and type of place of residence for
all the three surveys is presented at Table V.1.5.2. Results show that, except the West
region at TDHS-2003 and at the rural data at TDHS-1998 the median age of first
birth at age 40-44 and 45-49 are same or the last age groups is higher than the 40-44
age group. At these two situations the median ages are 1 year older at 45-49 age
group than 40-44.
143
Table V.1.5.1. Average Number of Children Ever Born by Age of Mother by
Region and Type of Place of Residence, TDHS 1993, 1998 and 2003
Age of Woman
15-19 20-24 25-29 30-34 35-39 40-44 45-49 Total
Region
West
1993 0.5 1.1 1.8 2.4 2.9 3.4 3.8 2.4
1998 0.1 0.6 1.4 2.1 2.6 3.2 3.7 1.7
2003 0.4 1.0 1.6 2.2 2.5 2.9 3.0 2.2
South
1993 0.5 1.4 2.1 3.1 3.9 4.6 5.3 3.1
1998 0.1 0.7 1.5 2.6 3.4 4.1 4.9 2.1
2003 0.8 1.2 1.9 2.6 3.1 3.7 4.1 2.7
Central
1993 0.5 1.3 2.4 3.0 4.1 4.5 5.1 3.1
1998 0.1 0.8 1.7 2.6 3.1 3.6 4.2 2.0
2003 0.6 1.2 1.7 2.5 2.9 3.3 3.7 2.5
North
1993 0.5 1.2 2.1 3.2 4.0 4.9 4.8 3.0
1998 0.0 0.8 1.9 2.7 3.1 4.2 4.6 2.2
2003 0.5 1.2 1.9 2.5 3.1 3.4 3.6 2.7
East
1993 0.6 1.8 3.1 4.6 6.3 7.1 7.8 4.3
1998 0.1 1.1 2.6 3.9 5.3 6.6 7.4 2.7
2003 0.7 1.6 2.8 3.8 4.8 5.9 6.6 3.8
Type of Place of Residence
Urban
1993 0.5 1.2 2.0 2.8 3.5 4.0 4.3 2.7
1998 0.1 0.7 1.5 2.4 3.0 3.5 4.2 1.9
2003 0.5 1.1 1.8 2.4 2.8 3.3 3.4 2.4
Rural
1993 0.5 1.5 2.6 3.7 4.6 5.5 5.9 3.6
1998 0.1 0.9 2.1 3.0 3.8 4.9 5.2 2.3
2003 0.7 1.3 2.4 3.0 3.6 4.3 4.5 3.1
Total
1993 0.5 1.3 2.2 3.1 3.9 4.5 4.9 3.0
1998 0.1 0.8 1.7 2.6 3.3 4.0 4.5 2.0
2003 0.6 1.2 1.9 2.6 3.1 3.5 3.7 2.6
144
Table V.1.5.2. Median Age at First Birth by Age of Woman at the Time of
Survey by Region and Type of Place of Residence, TDHS 1993, 1998 and 2003
Age of Woman
20-24 25-29 30-34 35-39 40-44 45-49 Total
Region
West
1993 20 20 20 20 20 20 20
1998 20 21 21 20 20 20 20
2003 19 21 21 21 20 21 21
South
1993 19 20 20 20 20 20 20
1998 19 21 21 21 20 19 20
2003 19 21 21 21 21 21 21
Central
1993 19 20 20 19 19 20 19
1998 19 21 20 20 20 20 20
2003 19 20 21 20 20 20 20
North
1993 20 21 19 20 20 19 20
1998 20 21 21 20 20 20 20
2003 20 21 20 21 21 20 20
East
1993 18 19 19 19 20 20 19
1998 18 19 19 19 19 19 19
2003 19 20 20 19 19 19 19
Type of Place of Residence
Urban
1993 20 20 20 20 20 20 20
1998 19 21 21 20 20 20 20
2003 19 21 21 21 20 21 21
Rural
1993 19 20 19 19 20 20 19
1998 19 20 20 20 19 20 20
2003 19 20 20 20 20 20 20
Total
1993 19 20 20 20 20 20 20
1998 19 21 21 20 20 20 20
2003 19 21 21 21 20 20 20
145
V.2. The Quality of Death Related Data
The birth history module of the DHS questionnaire collects information on all live
births ignoring the current status of living of the children. The birth information of all
children whether died or living at the time of survey is aimed to be collected
accurately and complete. For the dead children additional questions were asked to
collect information on the exact dead date. At this study, the quality of the death
related data is aimed to be assessed focusing on the birth and dead death of the dead
children.
V.2.1. Date of Birth Data
The assessment of the quality of the death related data starts with the date of birth
data comparison of the living and dead children. Table V.2.1.1. illustrates the
percentage of births with incomplete information on date of birth by survival status
by region and type of place of residence, TDHS 1993, 1998 and 2003. The results
indicate that there is a clear difference between the children dead and alive in respect
to the completeness of the birth information. It is seen that for all the three surveys,
the completeness of the information of dead children’s birth dates are less complete
than the living children. While the complete information on birth date of living
children is 97.6 % at TDHS-1993, for the same survey 88.1 of the dead children’s
birth date information is complete. At TDHS-1998, the 88.1 of the living children’s
birth date is reported complete, but the completeness of the birth date is 62.3 %
among the dead children with the gap increased to 25.8 %. Although the
completeness rates are still below the levels of TDHS-1993, at TDHS-2003 the
completeness of the birth date of living and dead children are respectively 94.4 %
and 77.0 %.
The completeness problem of dead children is higher in rural areas except TDHS-
1998. At this year, the completeness of the birth information of dead children is
calculated 59.8 % and % 89.9 for living children. Both at urban and rural areas, the
146
gap between the living and dead children is around 9 percent at TDHS-1993. The
rural data of TDHS-2003 have less completion on the birth dates of both living and
children as compared to TDHS-1993. The gap between the dead and living children’s
information seems high at rural areas for this survey.
The South region has the highest completion rates of dead children’s birth dates as
compared to other regions for the first two TDHSs. The lowest rates are seen at
North region (80.1 %) at TDHS-1993, and East region at TDHS-1998 and TDHS-
2003 (respectively, 52.7 % and 61.7%). The gap between the living children and the
dead in terms of completeness of the date of birth data is highest at West region at
TDHS-1998 with 32.2%. Except the region South, for all regions the gap of
completeness is above 23 %. The highest gap among the dead and living children’s
birth date information is seen at East Region at TDHS-2003.
Table V.2.1.2.shows the percentage of births with incomplete information on date of
birth by survival status by age and education level of women, TDHS-1993, 1998 and
2003. The women aged 15-19 gives complete birth date information both for the
living and the dead children. With small fluctuations, the completeness of the birth
date information of the dead children decreases with the age of the mother. Women
at older ages seem have problem of remembering especially the birth dates of their
dead children completely. TDHS-1993 results show that, women between the ages
20-34 give complete information for more than 90% of the births regarding the
survival status of the children. The lowest completeness of dead children is seen at
age group 40-44 with nearly 85%.
The education of the women seems have relationship between the completeness of
the birth date information. The percent of complete information among dead children
in terms of birth date increases when the level of mother’s education increases.
While 86.2%, 49.9 % and 65.9 % of the dead children’s birth information is
complete among women with no education or did not complete the primary school at
TDHS-1993, TDHS-1998 and TDHS-2003 respectively, 95.7, 82.3 and 99.0 of the
birth dates of the dead children is recorded completely for the same years among the
147
women having secondary or higher education. The difference in terms of reporting
the complete information on birth information of the dead children between the
women did not have any education or did not complete the primary level the
secondary or higher educated increased from 9.5 % at TDHS-1993 to 33.1 % at
TDHS-2003.
148
Table V.2.1.1. Percentage of Births with Incomplete Information on Date of
Birth by Survival Status by Region and Type of Place of Residence, TDHS 1993,
1998 and 2003.
Living Children Dead Children
No
information
missing
Any
Information
missing
Month Only
Missing
Yr., Month
and Age are
Missing
No
information
missing
Any
Information
missing
Month Only
Missing
Yr., Month
and Age are
Missing
All
Births
Region
West
1993 98.7 0.0 1.3 0.0 88.5 8.1 0.0 3.4 5688
1998 93.5 0.7 5.8 0.0 61.2 35.1 0.0 3.7 5414
2003 97.5 0.3 2.2 0.0 84.6 13.4 0.0 2.0 7232
South
1993 97.6 0.1 2.2 0.0 91.9 7.7 0.0 0.5 3096
1998 90.7 0.3 8.9 0.1 75.1 24.6 0.0 0.3 2579
2003 92.5 2.2 5.3 0.0 74.4 22.2 0.0 3.4 2766
Central
1993 98.1 0.1 1.8 0.0 88.0 11.5 0.0 0.5 4668
1998 90.9 1.2 7.7 0.2 67.6 28.9 0.0 3.5 3895
2003 97.5 0.3 2.1 0.0 87.1 12.4 0.0 0.5 4691
North
1993 94.0 1.6 4.4 0.0 80.1 17.5 0.0 2.5 1851
1998 87.7 2.6 9.7 0.0 63.5 32.0 0.0 4.5 1500
2003 96.3 2.1 1.6 0.0 79.4 13.0 0.0 7.6 1583
East
1993 97.0 0.4 2.5 0.0 88.8 10.3 0.0 0.8 4524
1998 75.8 2.0 22.2 0.0 52.7 39.6 0.0 7.7 3821
2003 86.8 3.7 9.4 0.0 61.7 32.1 0.0 6.2 4901
Type of Place of Residence
Urban
1993 98.3 0.1 1.6 0.0 89.2 9.2 0.0 1.7 11332
1998 89.9 0.7 9.4 0.0 59.8 36.9 0.0 3.3 10644
2003 96.0 0.9 3.1 0.0 80.3 16.8 0.0 2.9 13986
Rural
1993 96.6 0.5 2.9 0.0 87.2 11.5 0.0 1.3 8495
1998 85.2 2.0 12.7 0.1 65.4 28.8 0.0 5.8 6565
2003 91.1 2.5 6.4 0.0 72.4 23.4 0.0 4.2 7187
Total
1993 97.6 0.3 2.1 0.0 88.1 10.4 0.0 1.5 19827
1998 88.1 1.2 10.6 0.1 62.3 33.3 0.0 4.4 17209
2003 94.4 1.5 4.2 0.0 77.0 19.6 0.0 3.4 21173
149
Table V.2.1.2. Percentage of Births with Incomplete Information on Date of
Birth by Survival Status by Demographic Characteristics of Woman, TDHS
1993, 1998 and 2003.
Living Children Dead Children
No
information
missing
Any
Information
missing
Month Only
Missing
Yr., Month
and Age are
Missing
No
information
missing
Any
Information
missing
Month Only
Missing
Yr., Month
and Age are
Missing
All
Births
Age
15-19
1993 100.0 0.0 0.0 0.0 100.0 0.0 0.0 0.0 177
1998 100.0 0.0 0.0 0.0 100.0 0.0 0.0 0.0 159
2003 100.0 0.0 0.0 0.0 100.0 0.0 0.0 0.0 140
20-24
1993 99.4 0.0 0.5 0.1 93.0 3.3 0.0 3.7 1387
1998 98.5 0.0 1.5 0.0 89.0 11.0 0.0 0.0 1190
2003 99.5 0.0 0.5 0.0 95.0 5.0 0.0 0.0 1249
25-29
1993 99.3 0.0 0.7 0.0 93.7 6.3 0.0 0.0 2705
1998 95.8 0.5 3.7 0.0 82.6 15.6 0.0 1.8 2384
2003 97.7 0.3 2.0 0.0 84.3 12.3 0.0 3.4 2846
30-34
1993 98.5 0.3 1.2 0.0 92.8 5.8 0.0 1.4 3966
1998 91.4 0.7 7.9 0.0 73.2 24.5 0.0 2.3 3110
2003 96.3 0.7 3.0 0.0 77.7 20.9 0.0 1.4 3879
35-39
1993 98.0 0.1 1.9 0.0 88.3 11.2 0.0 0.5 4176
1998 88.6 1.2 10.2 0.0 60.6 32.7 0.0 6.7 3537
2003 94.6 1.9 3.5 0.0 73.8 22.4 0.0 3.8 4339
40-44
1993 95.8 0.7 3.5 0.0 84.6 12.3 0.0 3.1 4070
1998 82.3 1.9 15.8 0.0 59.8 33.7 0.0 6.5 3501
2003 91.9 2.4 5.6 0.0 76.0 18.8 0.0 5.2 4719
45-49
1993 95.5 0.4 4.1 0.0 85.7 13.4 0.0 0.9 3347
1998 79.9 2.0 17.8 0.3 51.0 45.4 0.0 3.6 3327
2003 90.3 2.0 7.6 0.0 76.0 21.2 0.0 2.9 4001
150
Table V.2.1.2. Percentage of Births with Incomplete Information on Date of
Birth by Survival Status by Demographic Characteristics of Woman, TDHS
1993, 1998 and 2003. (Continued)
Living Children Dead Children
No
information
missing
Any
Information
missing
Month Only
Missing
Yr., Month
and Age are
Missing
No
information
missing
Any
Information
missing
Month Only
Missing
Yr., Month
and Age are
Missing
All
Births
Education
No educ/Pri. Inc.
1993 95.8 0.6 3.6 0.0 86.2 11.8 0.0 2.0 10042
1998 76.5 2.5 20.9 0.1 49.9 43.8 0.0 6.3 7378
2003 86.3 3.8 9.8 0.0 65.9 28.4 0.0 5.8 7435
Primary
1993 99.1 0.0 0.9 0.0 91.7 7.8 0.0 0.4 8776
1998 95.7 0.3 3.9 0.1 80.4 17.9 0.0 1.7 8603
2003 98.2 0.3 1.5 0.0 87.8 11.1 0.0 1.1 11732
Secondary +
1993 99.9 0.0 0.1 0.0 95.7 4.3 0.0 0.0 1009
1998 99.8 0.0 0.2 0.0 82.3 17.7 0.0 0.0 1227
2003 99.9 0.0 0.1 0.0 99.0 1.0 0.0 0.0 2005
Total
1993 97.6 0.3 2.1 0.0 88.1 10.4 0.0 1.5 19827
1998 88.1 1.2 10.6 0.1 62.3 33.3 0.0 4.4 17209
2003 94.4 1.5 4.2 0.0 77.0 19.6 0.0 3.4 21173
V.2.2. Age at Death Data
The standard DHS questionnaire allows the interviewer to record the age at death
information either by days, months or year by circling the appropriate unit according
to the response of the women. The two variables at the standard recode DHS data
keeps the coded version at the field and the recoded version of the death age
information in months. This section of the study deals with the quality of the age at
death information. The completeness of the birth data information is studied at the
previous sections of the study. The completeness of the age at death information
gives the level of imputation problem for the age at death which will result in
changes in mortality estimations.
151
Table V.2.2.1. shows the distribution of the deaths among children by calendar
period in which the birth occurred by type of place of residence and region where the
women lives at the time of survey. Among the three surveys, TDHS-1998 seems
more problematic in terms of completeness of the age at death data especially for the
5 years preceding the survey. In general, 6.7 % of the deaths were reported as
incomplete at this survey. Incompleteness is 1.4 at TDHS-1993 and 2.5 % at TDHS-
2003. The quality of the information decreases especially at TDHS-1993 for the
calendar year periods of 10 or more. While the completeness of the age at death is
same at the first two surveys at 0-24 years, at the last TDHS, the quality of the
overall age at death data seems lowest for the same period.
TDHS-1998 results show that %8.3 of the rural deaths and 5.2 % of urban deaths are
incomplete for the births happened within 5 year period before the survey. The
quality of the data age at death is better in rural at TDHS-1993 and TDHS-2003 for
the same period. However, regarding the period of 0-24, the urban data is better in
terms of completeness of the age at death data.
For the births of 0-5 period preceding the survey, the age at death information is
complete at West and North Regions for all the three surveys. At TDHS-2003 at
North Region for the same period the age at death data is also completed. The same
situation is seen at TDHS-1993 for East Region. At TDHS-2003 all the death
information seems complete at West region for all calendar year periods of births
preceding the survey. The highest quality of the data is seen at this region for TDHS-
2003 and TDHS-1993. For TDHS-1998 the highest completeness percent is seen at
North region. The lowest quality age at death data, on the other hand, is seen at
North, East and Central regions respectively at TDHS-1993, TDHS-1998 and TDHS-
2003.
152
Table V.2.2.1. Percentage of Deaths with Incomplete Information on Age at
Death by Calendar Period in Which the Birth Occurred by Region and Type of
Place of Residence, TDHS 1993, 1998 and 2003.
Calendar-Year Period Preceding Survey
0-4 5-9 11-14 15-19 20-24 0-24
Region
West
1993 0.0 0.0 1.1 1.9 1.1 1.0
1998 0.0 6.7 4.0 3.4 4.0 3.9
2003 0.0 0.0 0.0 0.0 0.0 0.0
South
1993 2.3 5.3 0.0 2.1 6.2 3.1
1998 10.8 0.0 7.9 0.0 3.3 3.5
2003 14.2 3.6 7.6 0.0 0.0 3.8
Central
1993 4.1 0.0 4.6 2.9 4.7 3.3
1998 1.8 0.0 0.0 1.9 1.9 1.2
2003 0.0 0.0 2.4 12.7 2.8 4.6
North
1993 0.0 3.3 7.8 7.7 7.3 6.2
1998 0.0 1.7 0.0 2.3 0.0 1.0
2003 0.0 2.2 0.0 5.3 0.0 1.6
East
1993 0.0 3.2 4.9 7.2 8.3 5.0
1998 13.3 1.9 3.7 4.5 6.9 5.8
2003 2.4 1.6 0.8 1.5 1.2 1.4
Type of Place of Residence
Urban
1993 2.7 0.8 2.5 3.6 4.6 2.9
1998 5.2 4.6 3.2 2.6 2.3 3.2
2003 3.1 1.0 1.2 1.6 1.4 1.5
Rural
1993 0.0 3.3 4.6 4.5 5.1 4.0
1998 8.3 0.0 2.4 3.2 5.7 3.7
2003 1.7 1.6 1.9 6.4 0.4 2.5
Total
1993 1.4 2.1 3.6 4.1 4.9 3.5
1998 6.7 2.6 2.8 2.8 3.7 3.5
2003 2.5 1.3 1.5 3.6 1.0 1.9
153
Table V.2.2.2. Percentage of Deaths with Incomplete Information on Age at
Death by Calendar Period in Which the Birth Occurred by Demographic
Characteristics of Woman, TDHS 1993, 1998 and 2003.
Calendar-Year Period Preceding Survey
0-4 5-9 11-14 15-19 20-24 0-24
Age
15-19
1993 0.0 - - - - 0.0
1998 0.0 0.0 - - - 0.0
2003 0.0 - - - - 0.0
20-24
1993 0.0 0.0 50.0 - - 1.3
1998 10.4 0.0 - - - 7.7
2003 0.0 0.0 0.0 - - 0.0
25-29
1993 3.9 3.3 1.4 - - 3.0
1998 8.8 4.7 0.0 0.0 - 5.1
2003 0.0 2.9 0.0 - - 1.4
30-34
1993 0.0 1.9 4.3 0.0 0.0 2.3
1998 0.0 3.9 0.9 6.4 0.0 2.8
2003 4.2 0.6 2.2 9.6 - 2.7
35-39
1993 4.7 2.4 2.9 3.5 6.0 3.6
1998 0.0 0.0 3.9 0.0 2.0 1.3
2003 8.3 1.6 0.7 4.4 3.5 3.0
40-44
1993 0.0 0.0 5.1 5.5 4.8 4.5
1998 9.5 0.0 6.1 2.5 4.7 4.0
2003 11.5 0.0 0.7 1.1 0.8 1.0
45-49
1993 0.0 4.0 1.6 5.2 4.6 4.1
1998 23.2 0.0 2.1 5.9 3.5 4.1
2003 0.0 0.0 4.8 4.2 0.3 1.7
Total
1993 1.4 2.1 3.6 4.1 4.9 3.5
1998 6.7 2.6 2.8 2.8 3.7 3.5
2003 2.5 1.3 1.5 3.6 1.0 1.9
154
Table V.2.2.2. Percentage of Deaths with Incomplete Information on Age at
Death by Calendar Period in Which the Birth Occurred by Demographic
Characteristics of Woman, TDHS 1993, 1998 and 2003. (Continued)
Calendar-Year Period Preceding Survey
0-4 5-9 11-14 15-19 20-24 0-24
Education
No educ/Pri. Inc.
1993 1.9 2.8 4.4 5.2 6.2 4.5
1998 11.7 3.2 3.4 3.5 5.0 4.5
2003 6.0 1.4 2.0 2.8 1.0 2.2
Primary
1993 1.0 1.1 2.3 2.1 1.7 1.7
1998 3.9 2.2 1.9 0.5 0.9 1.8
2003 0.0 1.0 0.9 5.0 1.0 1.8
Secondary +
1993 0.0 0.0 0.0 0.0 0.0 0.0
1998 0.0 0.0 0.0 15.8 0.0 5.8
2003 0.0 2.6 0.0 0.0 0.0 0.7
Total
1993 1.4 2.1 3.6 4.1 4.9 3.5
1998 6.7 2.6 2.8 2.8 3.7 3.5
2003 2.5 1.3 1.5 3.6 1.0 1.9
Table V.2.2.2. illustrates the incompleteness of the age at death information by
calendar year period preceding the survey with controlling the age and educational
status of the mother. Results show that, there is no clear relationship between the age
of women and the completeness of the information. For different surveys, at different
ages complete information on the age at death is seen. The TDHS-1998 results
especially for the age groups 20-24 and 25-29 disorders the trends at the
completeness levels. On the other hand, table indicates that, when the education level
of the mother increases the completeness of the age at death information increases.
Except the TDHS-1998 data of the period of 15-19 years preceding the survey,
women having secondary or more education nearly gives complete information for
all calendar year periods preceding the surveys. The highest incompleteness is seen
among the women who did not have any education at all or did not complete the
primary school.
155
The defect in the age at death information gives the problem in recording the data at
the field. The number of deaths and the deaths without complete information is
presented at Table V.2.2.3. with the detailed information of the defect. In TDHS-
1993 2418 deaths are mentioned by the mothers while filling the birth history. 79 of
the deaths’ information are not complete. 82.3 % of these deaths were noticed as
inconsistent with other answers at the questionnaire and imputed during data entry.
While at only one death, some information is missing; all information is missing at
13 deaths at this survey. At TDHS-1998 and TDHS-2003 for nearly 30 % of the
deaths reported incompletely it is seen that all information is missing. The
inconsistent response problem is seen 58.9 % and 70% of the deaths with incomplete
age at death information at TDHS-1993 and TDHS-2003 respectively.
The data of rural areas seems more problematic in terms of inconsistent response
problem at TDHS-1993 and TDHS-2003.88.6 % and 66.7 % of the deaths reported
incomplete at these surveys are inconsistent with the other answers. At TDHS-2003,
for rural areas, the problem of all information missing reaches to 46.7 % of the
deaths lacking complete age at death information. At this survey, the inconsistent
response problem reaches to its highest levels at urban areas.
At West region, all the information of the deaths mentioned is calculated as complete
at TDHS-2003. At region North only 1 out of 159 and 117 deaths are reported
incomplete at TDHS-1998 and TDHS-2003 respectively. At TDHS-1993 and TDHS-
1998 the highest number of deaths reported incomplete is seen at East region with 30
and 25 deaths in that order. 93.3 % and 72.0 % of these deaths were mentioned as
inconsistent with other information and imputed at the data entry process. At TDHs-
2003 the highest incomplete information at age at death data is seen at Central region
with 16 deaths. While for 43.7 % of the incomplete deaths, all information is
missing, 56.3 % of these deaths’ information is inconsistent with other data.
156
Table V.2.2.3. Total Reported Deaths and the Number of Deaths with
Incomplete Age at Death Information by Type of Defect in Information by
Region and Type of Place of Residence, TDHS 1993, 1998 and 2003.
Total
Reported
Deaths
Deaths Lacking
Complete Age at
Death Information
Defect in Age at Death Information
All Information
Missing
Some Information
Missing
Inconsisten
t Response
Region
West
1993 580 6 2 0 4
1998 505 16 4 4 7
2003 511 0 0 0 0
South
1993 322 9 0 1 8
1998 198 5 3 0 3
2003 187 5 1 0 4
Central
1993 653 21 6 0 15
1998 429 8 5 0 3
2003 387 16 7 0 9
North
1993 220 12 2 0 10
1998 159 1 0 0 1
2003 117 1 0 0 1
East
1993 642 30 2 0 28
1998 462 25 7 0 18
2003 494 7 1 0 6
Type of Place of Residence
Urban
1993 1136 35 9 1 25
1998 979 31 10 4 17
2003 985 15 2 0 12
Rural
1993 1281 44 4 0 39
1998 773 24 8 0 16
2003 711 15 7 0 8
Total
1993 2418 79 13 1 65
1998 1753 56 18 4 33
2003 1697 30 9 0 21
157
The number of deaths mentioned at age group 15-19 is very few as they are at the
beginning of the reproductive ages. It is known that births are also very few at these
ages. All the deaths reported at this age group are complete in terms of the
information on age at death. Even till to age 35 and above very few deaths are
mentioned as incomplete. As the births increase in numbers with the age of women,
the number of deaths also increases. The number of deaths in which the age of death
information is not complete increases with the age of women. The deaths of all the
children regardless of birth date are collected at the birth history. A woman at the last
reproductive age group reports the death of her child even she gave birth at the first
ages of reproduction. Therefore the level of completeness is decreasing at the older
age groups. .
Table V.2.2.4. also indicates the completeness of the information and the defect of
information at the incomplete age at death data according to the education of the
women interviewed. The highest numbers of deaths are reported by the women with
no education at all or did not complete the primary education. The incomplete age at
death information at this group of women is mainly sourced from the inconsistent
data with other information. Although, the number of deaths mentioned by the
women having secondary or higher education is increasing among the surveys; only
2 of the reported deaths at TDHS-1998 are lacking complete age at death
information. No deaths are incomplete at TDHS-1993 and TDHS-2003 at the data
reported by the women at this level of education.
158
Table V.2.2.4. Total Reported Deaths and the Number of Deaths with
Incomplete Age at Death Information by Type of Defect in Information by
Demographic Characteristics of Woman, TDHS 1993, 1998 and 2003
Defect in Age at Death Information
Total
Reported
Deaths
Deaths Lacking
Complete Age at
Death Information
All
Information
Missing
Some
Information
Missing
Inconsistent
Response
Age
15-19
1993 16 0 0 0 0
1998 10 0 0 0 0
2003 6 0 0 0 0
20-24
1993 91 1 0 0 1
1998 67 5 2 0 4
2003 53 0 0 0 0
25-29
1993 202 6 0 1 5
1998 145 7 0 0 7
2003 118 2 0 0 2
30-34
1993 387 9 0 0 9
1998 241 7 2 2 2
2003 232 6 2 0 3
35-39
1993 485 17 4 0 12
1998 350 4 2 0 2
2003 347 10 6 0 4
40-44
1993 607 25 5 0 20
1998 397 15 5 0 10
2003 456 4 0 0 4
45-49
1993 630 21 4 0 17
1998 543 18 8 2 8
2003 485 8 0 0 8
159
Table V.2.2.4. Total Reported Deaths and the Number of Deaths with
Incomplete Age at Death Information by Type of Defect in Information by
Demographic Characteristics of Woman, TDHS 1993, 1998 and 2003.
(Continued)
Defect in Age at Death Information
Total
Reported
Deaths
Deaths Lacking
Complete Age at
Death Information
All
Information
Missing
Some
Information
Missing
Inconsistent
Response
Education
No educ/Pri. Inc.
1993 1592 66 8 1 58
1998 1043 42 14 4 24
2003 865 15 2 0 13
Primary
1993 796 12 5 0 7
1998 671 12 3 0 9
2003 773 14 7 0 7
Secondary +
1993 29 0 0 0 0
1998 38 2 2 0 0
2003 59 0 0 0 0
Total
1993 2418 79 13 1 65
1998 1753 56 18 4 33
2003 1697 30 9 0 21
V.2.3. Accuracy of the data:
Heaping the month of the death of children to 12 is a common problem of DHS.
During the training, interviewers were told to probe the answer 1 year for age at
death, whether it is exactly 12 months or not. If there is a heaping on month 12, DHS
program did not develop a system to distribute the excess deaths to other months.
The simple solution to overcome this problem is to stress the importance of to probe
the 1 year answers. Heaping on 12th month has a direct effect on infant and child
160
mortality rates. However, like any other DHS survey reports, TDHS survey reports
are not adjusted for the heaping while estimating the infant and child mortality rates.
The extent of the problem is studied at this study with some socio-demographic
variables of the women.
Table V.2.3.1. illustrates the index of heaping of death at 12 months of age by region
and type of place of residence. If there is no heaping at the data, the index score must
be 1.0. Any number above 1 indicates a level of heaping for 12th month. If the score
is calculated less than 1.0, the number of deaths reported at month 12 must be less
than the neighboring months. All Index results indicate that the highest heaping is
seen at TDHS-1998 with 8.6 of an index score. The lowest heaping is seen at TDHS-
2003 with a score of 5.The lowest level also indicates a heaping on month 12. This
indicates 5 times high preference of this month than the neighboring months. At
TDHS-1998 the index is calculated as 6.8.
The heaping at rural areas is higher than the urban areas for all surveys. The heaping
is highest atTDHS-1998 (9.9) at rural areas. Ignoring the results of TDHS-1998 the
heaping seems decreasing, at TDHSs. On the other hand, the lowest heaping at
month 12 at TDHS-1993 and TDHS-2003 is seen at West region with the index
scores 2.9 and 3.1 respectively. At TDHS-1998, while the lowest heaping is seen at
Central region (4.1), the highest score among all regions is seen at this survey at
North region with an index score of 26.2. The highest heaping is seen at East region
(18.7) for TDHS-1993 and South region (13.9) at TDHS-2003.
161
Table V.2.3.1. Index of Heaping of Deaths at Twelve Months of Age by Region
and Type of Place of Residence, TDHS 1993, 1998 and 2003
Month of Death Index of Heaping
10 11 12 13 14 at month 12
Region
West
1993 5 17 20 1 4 2.9
1998 3 7 21 0 2 7.2
2003 0 6 8 1 4 3.1
South
1993 3 6 19 1 2 6.0
1998 2 4 12 1 1 5.6
2003 0 1 11 2 1 13.9
Central
1993 11 8 39 0 2 7.3
1998 3 6 13 2 2 4.1
2003 4 6 17 2 0 5.7
North
1993 2 7 12 1 0 4.5
1998 0 1 16 0 1 26.2
2003 2 3 7 0 1 4.9
East
1993 6 1 51 2 1 18.7
1998 3 6 36 0 1 13.9
2003 6 4 19 5 4 3.9
Type of Place of Residence
Urban
1993 14 17 58 2 4 6.2
1998 6 14 46 2 3 7.5
2003 8 13 34 5 6 4.3
Rural
1993 14 23 82 3 5 7.3
1998 6 10 51 1 4 9.9
2003 5 7 29 5 3 5.8
Total
1993 28 40 140 5 9 6.8
1998 12 24 97 2 7 8.6
2003 13 20 63 10 8 4.9
162
Table V.2.3.2. Index of Heaping of Deaths at Twelve Months of Age by
Demographic Characteristics of Woman, TDHS 1993, 1998 and 2003.
Month of Death Index of Heaping
10 11 12 13 14 at month 12
Age
15-19
1993 1 0 0 0 0 0.0
1998 0 0 0 0 0 -
2003 0 0 0 0 0 -
20-24
1993 1 1 3 0 0 6.9
1998 1 1 4 1 0 4.8
2003 1 0 0 0 1 0.0
25-29
1993 0 2 11 0 0 17.4
1998 1 3 8 0 1 5.7
2003 3 1 2 1 0 2.1
30-34
1993 4 11 22 1 3 5.0
1998 2 1 14 1 0 14.1
2003 1 1 9 2 2 6.2
35-39
1993 4 2 22 1 2 9.3
1998 2 6 15 0 1 6.4
2003 3 4 15 0 2 6.7
40-44
1993 15 13 35 1 3 4.3
1998 2 5 29 0 2 12.0
2003 1 8 16 3 1 5.0
45-49
1993 3 11 47 2 1 11.2
1998 3 8 28 1 2 8.2
2003 3 6 20 5 4 4.8
163
Table V.2.3.2. Index of Heaping of Deaths at Twelve Months of Age by
Demographic Characteristics of Woman, TDHS 1993, 1998 and 2003.
(Continued)
Month of Death Index of Heaping
10 11 12 13 14 at month 12
Education
No educ/Pri. Inc.
1993 20 26 102 5 5 7.3
1998 6 8 77 1 3 16.9
2003 7 9 44 9 3 6.3
Primary
1993 8 14 38 0 4 5.8
1998 3 14 21 1 3 3.8
2003 5 11 18 2 4 3.4
Secondary +
1993 0 1 0 0 0 0.0
1998 2 2 0 1 1 0.0
2003 0 0 1 0 2 1.3
Total
1993 28 40 140 5 9 6.8
1998 12 24 97 2 7 8.6
2003 13 20 63 10 8 4.9
Index of heaping calculated according to the age and education of women is
presented at Table V.2.3.2.. No relation with the age of mother and the level of
imputation is seen. The highest index scores are estimated at age 25-29 (17.4), 30-34
(14.1) and 35-39 (67) respectively for TDHS-1993, 1998 and 2003. The education
level of the women and the level of heaping are calculated at this section of the
study. Results show that when the level of education increases the level of heaping
for month 12 is decreasing. The highest heaping is seen among women have no
education or did not complete the primary level for all surveys. The highest index
score is calculated as 16.9 at TDHS-1998 among the woman with no education or not
completed primary level. In addition to the tables on the heaping at month 12 at the
164
date of death data, figures at Annex VIII.4. illustrates the heaping on month 12 at
date of death information by selected socio-demographic characteristics of woman.
V.3. The Impact of Data Quality on Demographic Rates
V.3.1. Fertility impact of data quality
At this section of the study the impact of boundary effects and the sleeping away
distortion at the data on Total Fertility Rate and Under 5 Mortality Rate is aimed to
be assessed. Simulations are constructed and used for different situations that is
related with the boundary effect and sleeping away exclusion of the women.
Simulations are used to measure the effect of both lower and upper boundaries
separately. Five simulations were used to assess the impact of exclusion of woman at
the household questionnaire.
Simulation 1: FLB0 (Lower Boundary)
Excluded women had an age-specific fertility of 0.0 (no births)
Simulation 2: FUB0 (Upper boundary)
Excluded women had an age-specific fertility of 0.0 (no births)
Simulation 3: FUB2 (Upper Boundary)
Excluded women had twice the age-specific fertility as included
Simulation 4: FSA75 (Sleeping away)
Excluded woman had 75 percent of the age-specific fertility of included women
Simulation 5: FSA125 (Sleeping away)
Excluded woman had 125 percent of the age-specific fertility of included women
165
Table V.3.1.1. indicates the results of simulations for TDHS-1993 at regional and
residential separation. The overall results show that while the boundary effect is
estimated high at upper boundary estimation, no effect on fertility rates is seen.
Likewise, the effect of sleeping away exclusion did not have impact on Total
Fertility Rates at TDHS-1993. Simulations, on the other hand, created a 0.01 increase
at TFR estimated for rural area. At the urban area, simulation where the women
excluded because of upper boundary effect having no births resulted in only a 0.01
decrease at TFR. For all other simulations, nothing has changed at TFR at urban area.
The scenario where the women excluded because of the lower boundary effect
having no births create a 0.04 and 0.03 decrease at West and North region
respectively. Results also show that, if the women excluded from the last eligible age
group had no births, TFR at regions West, North and East decreases with a negligible
number of births. However, at the simulation where the women estimated as
excluded by the upper boundary effect have twice Age Specific Fertility Rate
(ASFR) as excluded, the TFR calculated is 3 times high than the TFR at the report at
East region. This is mainly sourced from the continuation of births still at the last
eligible age group of women
166
Table V.3.1.1. Results of Simulations to Estimate the Effect of Lower Boundary,
Upper Boundary and “Sleeping Away” Exclusions on Total Fertility Rate,
TDHS 1993
Actual
TFR
Total Fertility Rate
FLB0 FUB0 FUB2 FSA75 FSA125
Region
West 1.93 1.89 1.92 1.92 1.92 1.93
South 2.31 2.31 2.31 2.31 2.31 2.32
Central 2.44 2.44 2.44 2.44 2.44 2.44
North 2.76 2.73 2.75 2.75 2.75 2.76
East 4.07 4.07 4.05 12.80 4.07 4.07
Type of Place of Residence
Urban 2.31 2.31 2.30 2.31 2.31 2.31
Rural 2.87 2.88 2.88 2.88 2.88 2.88
Total 2.51 2.51 2.51 2.51 2.51 2.51
Table V.3.1.2. shows the result of simulations for TDHS-1998. It is seen that overall
TFR did not affected by the exclusions estimated by the boundary effect calculations
and sleeping away. TFR estimated at TDHS-1998 for urban areas was 2.39 and at
rural areas 3.08. The simulations FLB0 ad FUB0 where the women excluded at
lower boundary and upper boundary supposed to have no births did not make any
change at TFR for both residential areas. In addition the FUB2 and FSA75 and
FSA125 simulation results did not change TFR at urban. A 0.01 increase is seen after
the application of these simulations to the rural data. The simulation based on lower
boundary seems no effect on TFR at all regions. Simulation, FUB0 on the other
hand, only affected TFR at North region with a 0.03 decrease. The upper boundary
exclusion simulation where the excluded women are supposed to have twice ASFR
of the mother at the last reproductive ages, had effect only at East region where TFR
nearly doubled at this region to 8.39. At other regions, as there is no births recorded
167
for the age group 45-49 although the upper boundary effect is seen at some regions
TFR did not changed with the simulations. The effect of sleeping away exclusions on
TFR is also presented at Table IV.1.2.. Results show that the simulation where the
excluded women sourced by sleeping away exclusion having 75 percent of the agespecific
fertility of included women did not changed the TFR at all regions. The
excluded women by sleeping away factor are also simulated as they have 125 percent
of ASFR included women and this simulation resulted in a 0.01 increase only at
West and South regions.
Table V.3.1.2. Results of Simulations to Estimate the Effect of Lower Boundary,
Upper Boundary and “Sleeping Away” Exclusions on Total Fertility Rate,
TDHS 1998
Actual
TFR
Total Fertility Rate
FLB0 FUB0 FUB2 FSA75 FSA125
Region
West 2.03 2.03 2.03 2.03 2.03 2.04
South 2.55 2.55 2.55 2.55 2.55 2.56
Central 2.56 2.56 2.56 2.56 2.56 2.56
North 2.68 2.68 2.65 2.68 2.68 2.68
East 4.19 4.19 4.19 8.39 4.19 4.19
Type of Place of Residence
Urban 2.39 2.39 2.39 2.39 2.39 2.39
Rural 3.08 3.08 3.08 3.09 3.09 3.09
Total 2.61 2.61 2.61 2.61 2.61 2.61
168
The simulation results for TDHS-2003 on TFR are presented at Table V.1.3.3.
Although there is high level of exclusion of women at both lower and upper
boundary, it is seen that simulations on lower boundary effect have only effect on
TFR at urban areas and West region with 0.04 and 0.07 decrease respectively. The
upper boundary exclusion, on the other hand finds its reflection on a 0.01 decrease
with simulation FUB0 at Central region and rural data. The FUB2 simulation on the
other hand, increased TFR at Central region to 3.99. Negligible increase at TFR is
seen at East region and rural data with the simulation of FUB2. While sleeping away
exclusion with FSA75 simulation did not make any change at TFR, FSA125
simulation only increased TFR at East region with an increase of 0.01.
Table V.3.1.3. Results of Simulations to Estimate The Effect of Lower
Boundary, Upper Boundary and “Sleeping Away” Exclusions on Total Fertility
Rate, TDHS 2003
Actual
TFR
Total Fertility Rate
FLB0 FUB0 FUB2 FSA75 FSA125
Region
West 1.88 1.81 1.88 1.88 1.88 1.88
South 2.30 2.30 2.30 2.30 2.30 2.30
Central 1.86 1.86 1.85 3.99 1.86 1.86
North 1.94 1.94 1.94 1.94 1.94 1.95
East 3.65 3.65 3.65 3.66 3.65 3.66
Type of Place of Residence
Urban 2.06 2.02 2.06 2.06 2.06 2.06
Rural 2.65 2.65 2.64 2.66 2.65 2.65
Total 2.23 2.23 2.23 2.23 2.23 2.23
169
V.3.2. Mortality Impact of Data Quality
At this section of the study, the impact of data quality on mortality estimates at
TDHs in terms of exclusion of women from the household interview and the heaping
of deaths at twelve months of age at birth history section of the ever married women
questionnaire. Section starts with the impact of boundary effect and sleeping away
exclusion of Women on under-five mortality rate (U5MR) by applying some
simulations in which the women excluded have different levels of U5MR. The next
part of the section is the overall estimation of impact of heaping on month 12 at age
at death on infant mortality rate (IMR) and child mortality rate (CMR).
V.3.2.1.The Impact of Boundary Effect and Sleeping Away Exclusion on U5MR.
The sleeping away and boundary effect results are used at three simulations to
understand the effect of exclusion of women on under-five mortality rates (U5MR) at
each survey. Simulations are constructed and used for different situations that is
related with the boundary effect and sleeping away exclusion of the women. Three
simulations were used to assess the impact of exclusion of woman at the household
questionnaire.
Under the same ASFR as included women;
Simulation 1: (MLB150) Lower Boundary
Excluded children have 150 percent the rate under-five mortality by age of mother as
included children.
Simulation 2: (MUB150) Upper boundary
Excluded children have 150 percent the rate under-five mortality by age of mother as
included children.
170
Simulation 3: (MSA150) Sleeping away
Excluded children have 150 percent the rate under-five mortality by age of mother as
included children.
The results of the simulation on TDHS-1993 data is presented at Table V.3.2.1..
Results show that simulation results based on upper boundary exclusion have no
effect on U5MR. Ignorable increase on U5MR is seen with the simulations of lower
boundary and sleeping away exclusion. Upper boundary simulation resulted in no
change at urban and rural data.. At the urban data U5MR increased 0.02 and 0.05
with the MLB150 and MSA150 exclusions. At the rural data, while MSA150
resulted in no change, the simulation based on lower boundary exclusion increased
the U5MR 0.1 points. The regions which are not affected by lower boundary
exclusion are Central and East. At West simulation MLB150 resulted in 0.22
increase, this is 0.13 at North region. The highest effect of the lower boundary
exclusion simulated is at South region which resulted in an U5MR of 65.03. The
MUB150 simulation created difference at West and North region only. While U5MR
increased 0.77 points, at region North, the increment of U5MR is 1.19. The
simulation based on sleeping away exclusion changed U5MR at West with 0.01
points and at East with an increase of 1.21.
171
Table V.3.2.1.1. Results of Simulations to Estimate the Effect of Lower
Boundary, Upper Boundary and “Sleeping Away” Exclusions on Under-Five
Mortality Rate, TDHS 1993
Actual
U5MR
Under –five Mortality Rate
MLB150 MUB150 MSA150
Region
West 48.00 48.22 48.77 48.01
South 62.80 65.03 62.80 62.80
Central 69.22 69.22 69.22 69.22
North 49.52 49.65 50.71 49.52
East 70.37 70.37 70.37 71.58
Type of Place of Residence
Urban 50.50 50.52 50.50 50.55
Rural 76.43 76.53 76.43 76.43
Total 62.24 62.25 62.24 62.26
Table V.3.2.1.2. shows the results of simulations applied to TDHS-1998 data. U5MR
increased to 53.20 with the simulation based on lower boundary effect exclusion.
The upper boundary exclusion of women did not make any change for all regions
and rural and urban data. The only region that sleeping away exclusion seems
effective by the simulation MSA150 is the North region with a negligible increase of
0.04. In addition same simulation resulted in same quantity of increase at Rural data.
Simulation MLB150, on the other hand seems changed U5MR with different levels.
While U5MR increased 0.01 at rural areas, at North and Central region; the highest
increase is seen at 1.64 increase at South region.
172
Table V.3.2.1.2. Results of Simulations to Estimate the Effect of Lower
Boundary, Upper Boundary and “Sleeping Away” Exclusions on Under-Five
Mortality Rate, TDHS 1998
Actual
U5MR
Under –five Mortality Rate
MLB150 MUB150 MSA150
Region
West 38.25 39.58 38.25 38.25
South 43.04 44.68 43.04 43.04
Central 49.62 49.63 49.62 49.62
North 50.54 50.55 50.54 50.58
East 75.93 76.09 75.93 75.93
Type of Place of Residence
Urban 42.41 42.89 42.41 42.41
Rural 68.01 68.02 68.01 68.05
Total 53.05 53.20 53.05 53.05
The results of the simulations on U5MR at TDHS-2003 is shown at Table V.3.2.1.3..
Results indicate that the simulations based on upper boundary and sleeping away
exclusions did not make any noticeable change on the mortality rate. MLB150
simulation created difference in U5MR for all regions and urban and rural. However,
of simulation is seen highest at North region with a 2.77 increase than the actual
U5MR.. Moreover the effect of lower boundary excluded women who supposed to
heave 150 percent higher rate under-five mortality by age of mother as included
children resulted in a U5MR of 32.39 at South region. The impact of the exclusion of
women at lower boundary seems negligible both urban and rural U5MR. The upper
boundary exclusion have no effect on rural data and the data at regions except West.
The simulation MUB150 gives results of U5MR 0.25 and 0.15 points higher at West
region and urban areas respectively. Sleeping away exclusion seems have no effect
on U5MR or in insignificant amounts. U5MR increased at Central and East regions
173
with 0.32 and 0.21 increases in that order. While there is no difference at urban data,
simulation MSA150 resulted in 50.17 U5MR.
Table V.3.2.1.3. Results of Simulations to Estimate the Effect of Lower
Boundary, Upper Boundary and “Sleeping Away” Exclusions on Under-Five
Mortality Rate, TDHS 2003
Actual
U5MR
Under –five Mortality Rate
MLB150 MUB150 MSA150
Region
West 30.28 31.13 31.55 30.28
South 30.48 32.39 30.48 30.48
Central 32.78 33.30 32.78 33.10
North 47.76 50.53 47.76 47.76
East 48.59 48.63 48.59 48.80
Type of Place of Residence
Urban 30.41 30.49 30.66 30.41
Rural 50.07 50.30 50.07 50.17
Total 37.57 37.66 37.58 37.58
174
V.3.2.2. The Impact of Heaping of Deaths at Twelve Months of Age, on IMR and
CMR Estimates.
The impact of the heaping on IMR and CMR is estimated by reassigning the
proportion of deaths at 12 months to infancy period and calculating the mortality
rates according to the new distribution. The 25 percent of the excess deaths on month
12 are carried to the 0-11 month period for this calculation.
The results of the IMR and CMR are presented at Table V.3.2.2.1. for TDHS-1993,
1998 and 2003. Overall results indicate that the effect on heaping on month 12 at age
at death data is negligible. As the heaping of the deaths at the last 5 year preceding
the survey is small in numbers, the redistributing the excess deaths have less effect
on IMR and CMR estimated after the distribution. The fewest change on the rates are
seen at TDHS-2003. While IMR increased 0.4 %, CMR decreased 1.2 %. The
increase at TDHS-1998 is higher at IMR as compare to TDHS-1993 (1.4 and 1.0
percents respectively). On the other hand the decrease at CMR is higher at TDHS-
1993. While CMR decreased 6.0 percent at TDHS-1998, the percent change of the
same rate is 6.2 at TDHS-1993.
The IMR increase after redistributing the deaths on month 12 resulted in 2.2 increase
at rural data of TDHS-1998. This is the highest increase among all rural data. At the
same survey, CMR decreased 9.0 percent which is 2.6 and 6.8 points high than
TDHS-1993 and TDHS-2003.At the urban data, for all surveys the increase at IMR is
below 1 percent. The highest decrease is seen at urban data at TDHS-1993 with a 6.0
%.
The reassigning the heaping children on month 12 did not make changes more than 3
percent at all regions for all surveys. The highest increase is seen for TDHS-1993
data is at East region with 2.7 % . For TDHS-1998 and TDHS-2003, the highest
change is seen at South and Central regions respectively with 2.8 %and 2.9 %. After
the distribution of the deaths, there is no change seen at IMR at South and North
regions for TDHS-2003, and West region for TDHS-1998 and South region for
175
TDHS-1993. As the deaths of the children is happening at the early ages of the
children, as expected, the effect of redistributing the excess deaths at 12 months is
slightly more on CMR. The highest decrease is seen at North and East regions at
TDHS-1993 with a percent decrease of 13.7 and 12.6 respectively. For the same
survey while no change is seen at South region, the decrease on West and Central is
4.6 % and 3.3% respectively. At TDHS-1998, no difference is seen at West region in
terms of CMR. The order of the regions from highest change to lowest at this survey
is North, Couth, Central and East. The lowest decrease is seen at TDHS-2003 at
CMR as compared to other surveys. The CMR results of the regions South and North
did not change with the distribution of excess deaths of 12 months to earlier months.
The highest percent change at CMR is seen at Central region with 4.6.
The effect of heaping on 12th month on IMR and CMR is also estimated for 5 to 10
years preceding the surveys to compare the results of the heaping with the 5 year
period preceding the survey. Table VIII.5.1. at Annex VII.5. illustrates the results of
this estimation. The overall indicates that the heaping on the age at death data at 12th
months of the births of this period have more or less same amount of effect on the
rates. It is interesting that the region where the impact of the heaping seems
respectively high at five year prior to the survey decreased at 5 to 10 years preceding
the surveys. In contrast, the regions where the impacts are very low at 5 years
preceding the survey are found high at. 5 to 10 years prior to survey.
176
Table V.3.2.2.1. Estimates of Infant and Child Mortality for the Five year
Period Preceding the Survey, Adjusted for Heaping of Deaths at Twelve Months
of Age, by Region and Type of Place of Residence, TDHS 1993, 1998, 2003.
Infant Mortality (1q0) Child Mortality (4q1)
Unadjusted
Rate
Adjusted
Rate
Percent
Increase
Unadjusted
Rate
Adjusted
Rate
Percent
Decrease
Region
West
1993 42.66 42.91 0.6 5.59 5.33 4.6
1998 32.79 32.79 0.0 5.65 5.65 0.0
2003 22.09 22.13 0.2 8.38 8.33 0.5
South
1993 55.40 55.40 0.0 7.83 7.83 0.0
1998 32.70 33.62 2.8 10.69 9.75 8.8
2003 28.59 28.59 0.0 1.94 1.94 0.0
Central
1993 57.88 58.26 0.7 12.04 11.64 3.3
1998 41.26 41.93 1.6 8.72 8.04 7.8
2003 20.54 21.14 2.9 12.49 11.92 4.6
North
1993 44.20 44.93 1.7 5.57 4.81 13.7
1998 42.04 42.97 2.2 8.87 7.91 10.8
2003 33.96 33.96 0.0 14.28 14.28 0.0
East
1993 60.04 61.44 2.3 10.99 9.60 12.6
1998 61.52 62.50 1.6 15.35 14.37 6.4
2003 41.43 41.54 0.3 7.47 7.35 1.5
Type of Place of Residence
Urban
1993 44.04 44.44 0.9 6.76 6.35 6.0
1998 35.22 35.43 0.6 7.46 7.24 2.8
2003 23.43 23.52 0.4 7.15 7.06 1.3
Rural
1993 65.44 66.18 1.1 11.76 11.01 6.4
1998 55.01 56.21 2.2 13.75 12.51 9.0
2003 39.25 39.38 0.3 11.27 11.13 1.2
Total
1993 53.47 54.01 1.0 8.78 8.24 6.2
1998 43.25 43.84 1.4 9.80 9.21 6.0
2003 29.02 29.12 0.4 8.55 8.45 1.2
177
At this chapter of the study, the quality of the birth history section of the ever
married questionnaire is assessed. The birth and death date data has direct effect on
mortality and fertility rates estimated from TDHS. At this chapter by using different
tools of assessment the quality of birth and death information at birth history section
is evaluated. The overall quality of the data at birth history at all three TDHSs seems
not problematic. The information both at the birth and death date of the children is
nearly complete for the last 5 years. As the focus of the interviewers are collecting
complete data especially for the last 5 years preceding the survey, the birth and death
date data comes complete from the field. TDHS-1998 results draw inconsistent
results with the first and the last TDHS. For most of the estimations, it is hard to
mention a trend between the three surveys. TDHS-2003 results give high level of
complete information for both birth and death dates of the children. The data from
urban areas more complete than the data collected from rural in terms of birth date.
When the information is not complete the common situation is seen the complete
year and age but not the month. The mothers at earlier ages give complete birth date
information as the birth event occurs in maximum five year period. When the birth of
the children and the interview date is at far dates the data quality will be effected by
the remembrance problem. The mothers on the other hand having high level of
education are good at remembering their children’s birth dates completely. Education
seems has positively effect on the completeness of the birth date information. On the
other hand, the sex of the child and the time period that the interviewer has no effect
on the completeness of the information as the incompleteness of the data is directly
related with the mother’s response.
The displacement of the children from eligible ages for section 4 and section 5 at the
ever married woman questionnaire where various questions are asked to the mother
about the children born 5 year preceding the survey also studied at this chapter.
Although the level of displacement is ignorable in all the three DHSs, it is noticeable
at TDHS-1998 as compared to the first and the last TDHS. Displacement seems
decreased at TDHS-2003. While the displacement is seen at urban areas more clear
than rural, no regional difference is estimated. TDHS-1998 results seems distorting
178
the trends at the level of displacement on most of the socio-demographic
characteristics of the women studied.
The possible heaping problem at the birth data is also discussed. Heaping of age 6,
which is mainly a result of displacement is common for all level of estimations.
Moreover, heaping on age 8 and 13 is also common for three survey data. The survey
years ending with 3and 8 effects the heaping to these digits while calculating the age
of children by using the birth years which commonly ends with either 0 or 5. While
mentioning the birth year, the preference of digits 0 and 5 finds its reflection on the
age of children at age 8 or 13.
The miscalculation of the year because of the incomplete birth date information came
from the field is also the subject of this study. It is seen that, the imputed cases are
more vulnerable to miscalculation of year of birth problem than the not imputed
ones. When the percent of the imputed cases increase in a dataset, the possibility of
miscalculation increases. The problem seems more or less same at TDHS-1993 and
1998. Whereas, because of the reason that the fieldwork of TDHS-2003 is carried out
at the end of 2003 and at the first months of 2004, the techniques to estimate the
extent of the problem did not work at TDHS-2003. The quality of the TDHS-1998 is
questionable in terms of the high numbers of imputed cases at the birth date of the
children.
The coverage problem of the dead children is one of the data quality problems in
DHS especially at undeveloped or developing countries. Results indicate that except
a possibility of a coverage problem at TDHS-2003 for the age group of mother 45-
49, no coverage problem is seen for all other surveys and all other age groups.
The quality of the death related data at the birth history section is also discussed at
this chapter. The section starts with the evaluation of the completeness of the date of
birth data of the dead children with the comparison of survived one. For all surveys
the level of completeness of the dead children is lower than the living children. In
addition the incompleteness of the date of birth information of the dead children is
179
more common at rural areas. TDHS-1998 is an exception where the rural data is
more complete than the urban data as compared to two other surveys. The women
having secondary or higher education gave complete birth date information as
compared to women having lower education.
The completeness of the age at death data is also the subject matter of this study and
evaluated at this chapter. The overall completeness of the information is high in all
the three surveys. However, TDHS-1998 dataset seems more problematic when
compared to other two TDHSs in terms of the completeness of the date of death. This
survey results creates an inconsistency with TDHS-1993 and TDHS-2003. Without
considering the results of TDHS-1998, it is seen that, the urban data is more
complete at the birth date information than rural. While no relationship is seen
between the age of the mother and level of completeness at the date of death, is is
clear that the level of completeness increases the level of education of the woman at
all TDHSs. The defect at the death date data is mainly caused by the inconsistency of
the information with other data given by the mother. Although the problem of all
information missing seems increased at TDHS-1998 and TDHS-2003, the
inconsistency of the age at death data with other information is the main reason for
the defect at data of the same surveys.
The heaping on the month 12 at the death date of the children is also studied at this
chapter. Extent of the problem is evaluated with socio-demographic characteristic of
the respondent. Although the overall heaping level is very low, the highest heaping
among three surveys is seen at TDHS-1998. Rural data is estimated as being more
vulnerable to the heaping with regard to urban. There is a negative relation between
the education of the mother and the level of heaping. While the level of education of
mother increases, heaping on month 12 decreases.
At the last section of the chapter the impact of the data quality on demographic rates
are studied. The results of the exclusion of women at household data with the lower
and upper boundary and the sleeping away factors on total fertility rate and under 5
mortality rates. By constructing simulations for different possibilities of fertility and
180
mortality of children, the boundary effects and sleeping away exclusion is tested on
TFR and U5MR. The overall results show that the estimated problems at the
household data in terms of exclusion of woman at eligible ages seems either no or
negligible effect on TFR estimated for Turkey. One of the simulations to estimate the
effect of exclusion of women from the last eligible age group on fertility results in
high TFR only at East region.
The impact of the heaping on month 12 on infant and child mortality rates are also
discussed at the last section of the chapter. After redistributing the excess deaths on
month 12, new IMR and CMR are estimated for total, regions and urban and rural
residential areas. The results indicate that insignificant changes are seen at the actual
IMR and new estimations. The highest change is seen with a 2.9 change at TDHS-
2003 at Central region. While IMR at urban areas changed less at the first two
survey, for TDHS-2003 the change at IMR (although it is below 0.5 percent) is lower
at rural areas. The change at CMR on the other hand is also not in high levels. In
general, the change at CMR is below 10 % for all surveys. At some regions for
TDHS-1993 (North and East), and at region North for TDHS-1998 the difference
between the actual and the estimated CMR is above 10 %.
181
VI. CONCLUSION AND DISCUSSION
To obtain reliable and accurate information on the subject interested is valuable for
all scientists. To estimate a rare event’s frequency at nationwide, a large sample
should be selected. The DHS model is one of the programs that collect sociodemographic
data on various subjects including fertility and mortality. The success
of DHS program lies under the meeting the broad range of information needs,
producing results in a short time after the survey’s fieldwork is done, and having
standards starting from sampling procedures to report writing. Turkey carried out 3
surveys find place in DHS program. DHS core questionnaires are modified for the
needs of the country. The standard modules like household list at the household
questionnaire and basic characteristics of woman, reproduction, contraceptive use,
anthropometric measurement etc. is kept at the questionnaires used and additional
questions to supply the needs of the partner institutions and ministries in Turkey.
Hacettepe University Institute of Population Studies (HUIPS) is the implementing
institute in Turkey for the DHS. In 1993, 1998 and 2003 institute carried out Turkey
Demographic and Heath Surveys with a nationwide sample with both the technical
and economic support from the company responsible of DHS program all over the
world. HUIPS has a spectacular and well known history of carrying out demographic
surveys since 1968. After the establishment of the institute in 1967, institute held
demographic and health surveys conveniently with one of its aims.
Taking technical assistance from the program company is crucial in terms of the
overall quality of the survey and the data quality in particular. Collecting information
on a wide range of family planning and health topics resulted in the long and
complex questionnaires at TDHS. Therefore, the training period of the interviewers
and the field staff is 3 weeks. During the training, while giving education of filling
the questionnaires, health specialists give basic information about the reproduction
182
and other health related issues to the trainees. The last week of the training is mainly
left for pilot studies where the trainees are living the actual field at different parts of
the province. The quality of the training helps the quality of the data collected.
TDHS results are published by preliminary report where basic findings are presented
and the main report where detailed results are presented. The results of the TDHS are
frequently used by Ministry of Health, State Planning Organization and Turkish
Statistical Institute. Many important demographic and health indicators like Total
Fertility Rate, Infant, Child and Under Five Mortality Rates, Contraceptive
Prevalence Rate etc. are only estimated at these surveys. The official statistics
program based on the Statistics Law of Turkey No 5429, has been prepared for a 5-
year-period in order to determine the basic principles and standards dealing with the
production and dissemination of official statistics and to produce reliable, timely,
transparent and impartial data required at national and international level
(TURKSTAT, 2008) accepted and included some indicators on fertility and mortality
supplied by TDHS.
The data quality of the TDHSs is commonly accepted as reliable and acceptable.
However, in any large data-processing operation, it would be unrealistic to expect no
errors. Despite the importance and frequently used characteristics of the TDHS data,
the data quality of TDHS are not evaluated in a broad sense in terms of the quality of
the data having direct effect on fertility and mortality indicators. The first aim of this
study is to evaluate the data quality of TDHSs focusing on special information
effective on mortality and fertility rates. To assess the quality of TDHSs data to
expose the strength and weakness is important especially on evaluating the indicators
estimated form the surveys. All three TDHSs are evaluated and a comparison of
them is done to see the changes in the data quality.
Total Fertility Rate (TFR), Infant Mortality Rate (IMR), Child Mortality Rate (CMR)
and Under-5 Mortality Rate are the vital measures of the wellbeing of the children
and a good proxy indicator of the overall level of development of the country. The
variables have direct effect on the estimation of mortality of children and fertility is
183
the primary concern of this study and evaluated by using different methods and
indexes. The quality of the data evaluation starts with the quality of household data.
The eligibility of the women starts with the information at the household list.
Therefore, the starting point of the study is the quality of the age and usual residency
information. These two have direct effect of the number of women to whom the
individual questionnaire is applied. TDHS-1993, TDHS-1998 and TDHS-2003
member data is evaluated to understand the quality of data used for the eligibility of
the women.
Two sections of the ever married woman questionnaire are evaluated at this study.
Besides the quality of age information collected at the individual questionnaire, the
birth and death information of the children ever born is assessed. Women’s age have
direct effect on TFR. The Age Specific Fertility Rates (ASFR) are estimated for
conventional 5 year age groups of women which are summed to reach TFR. Birth
history section of the TDHSs give information about the birth and date dates and
ages of the live births. This study focuses on the data quality of birth history in terms
of these variables.
The second aim of this study is to assess the impact of the data quality of the selected
variables on the fertility and mortality rates estimated at TDHSs. The possible effect
of displacement of eligible women to not eligible age groups and sleeping away
exclusions of the eligible women is studied on TFR and U5MR by using different
simulations. The simulations which add the presumed excluded women to the
calculations of these two rates with different fertility and mortality characteristics.
The impact of the age heaping on age at death month 12 on infant and child mortality
rates are also evaluated at this study.
This study also aims to spot the common errors at TDHSs that need to be
concentrated on and develop suggestions for further field surveys. The results
estimated from TDHS experience will help researchers either studying the same
topics or having other interest areas while they are carrying on surveys.
184
Study starts with the introduction chapter where the purpose of the study is
discussed. The literature review chapter indicates the previous national and
international studies dealing on the quality of the data at surveys, particularly WFS
and DHS. The results of the quality of data of the variables used for the eligibility are
evaluated at Chapter IV. The household interview results are considered on the topic
of response rates which will create questions on the overall data quality of the survey
if it is under an accepted level. The quality of age data is also assessed in terms of
heaping, digit preference and boundary effect problem. The lower and upper
boundary effect results in displacement of women in eligible ages to non eligible
ages. In addition, the level of sleeping away exclusion of women at household data is
also studied. The age data at individual questionnaire is also evaluated at this chapter
of the study.
The quality of data collected at birth history section is evaluated in the division of
birth related data and death related data. The completeness and the displacement of
the birth dates of the children are brought into matter. In addition the age heaping
problem and the miscalculation of the birth problematic is evaluated. The last
assessments on the birth related data is done on the quality of coverage of live births.
The estimations of the quality of death related data includes the comparison of date
of birth of death and the surviving children. The age at death data is also evaluated in
terms of completeness of the information and the heaping on certain months
especially on the 12th month.
The implication of the quality of the data used for the eligibility of individual
questionnaire is assessed on TFR and U5MR. Using simulations based on the
displacement of women with boundary effects and the sleeping away exclusion, TFR
and U5MR are recalculated to measure the effect of these two data quality problems.
The implication of the heaping at month 12 on IMR and CMR is estimated. The
excess mortality at month 12 is redistributed to infancy and IMR and CMR is
recalculated.
185
The quality of data of TDHS is evaluated under two main heading. The first
assessment was done for the variables used for the eligibility of the women for
individual questionnaire and the age data at ever married woman questionnaire.
Evaluation starts with the household interview results. Although the response rates
are high for all the three surveys, results indicate that the response rates are
decreasing during the surveys. The rates are decreasing especially at the urban areas.
Coming from Hacettepe University and collaboration of the study with Ministry of
Health opens some doors for interview which cannot be opened by the interviewers
coming from private companies. Mainly the transportation of the field team is
supported by the official vehicles and this has a positive effect on nearly all areas.
The collaboration with Ministry of Health and using the ministry of health’s vehicles
during the field improves the response rates. For nearly all areas of Turkey, the name
of Hacettepe University brings the medical school in the mind of people. This also
have positive effect on the response rates. The high response rates are important for
East region. As this region has the highest fertility and mortality rates, to develop a
policy to this region in terms of health, high response rates will strong the basis of
the information collected at this area.
The overall quality of age reporting at household questionnaire seems at moderate
level. The heaping on ages ending with 0 and 5 is nearly universal. The demographic
surveys in Turkey is planned and applied to give intercensal demographic
estimations. Therefore since 1968, 8 surveys were carried out at the years ending
with 3 and 8. Although only the age not the birth month and year is asked at the
household questionnaire, the respondent will mention the birth year for some of the
members and this year of birth will vulnerable to heaping on 0 and 5. The heaping on
3 and will be a reflection of this kind of recoding the age of member calculated from
the birth year. The results indicate that, heaping is seen at both sexes but higher at
females.
Sex and Age are the two very important variables at the household data used for
deciding the eligibility of the women for the individual questionnaire. The data
quality of the female members at the household list is evaluated in details. Age and
186
sex ratios and Myers, Bachi, Whipple and United Nations indices for household data
for total, regions residential difference is estimated for female members. The overall
results indicate that the heaping problem at female age data is decreasing over the
surveys. For the first two TDHS the level of heaping estimated by Myers index is at
medium level, for TDHS-2003 this is estimated as low and acceptable. Over the
surveys both at urban and rural areas the quality of the data increases. To select the
best respondent to give quality data is important for the household interview. The
member of the household who will give complete and accurate information should be
selected. For nearly all surveys the percent of the situations where the age is
unknown is below 1 %. This indicates that, at the field, most of the time interviewers
are completing the household questionnaire with one of the members of the
household giving complete age information. The results which take the Myers,
Whipple and Bachi indices as basis for estimation indicate that the best information
is taken from either household head or her/his spouse. Less heaping is seen at the
data given by these people. In addition results show that the quality of the data is
very low when the respondent is above age 55. The best answers were taken from the
35-54 aged respondents at TDHS-1993 and TDHS-2003; and from 15-34 at TDHS-
1998.
The upper, lower and total boundary effect problems which are mainly sourced from
the displacement of eligible women out of eligible ages are evaluated for all TDHSs
at this study. Results indicate that there is no problem of displacement of women at
the lower boundary. However, the upper boundary effect, which has more effect on
the fertility and mortality indicators are quite common at TDHSs. Although the level
of the problem decreased from TDHS-1993 to TDHS-1998, the upper boundary
effect seems increased at the last TDHS. The regional difference is seen at the level
of problem. The highest problem is seen at East region where the highest fertility is
calculated for all three surveys. The exclusion of women at this region will have
effect on TFR more than the other regions. The lowest upper boundary effect is seen
at North region in general. Some selected background characteristics of the
respondent is also taken into consider while estimating the boundary effects. It is
187
estimated that the household head’s spouse who is mainly “female” gives low level
of boundary effect scores.
It is interesting that the lowest boundary effect problem is seen at TDHS-1998. This
may be a result of the training at this survey. While the quality of data at TDHSs is
increasing over the surveys, the increase at the problem of boundary exclusion will
be explained by the interviewer effect who will exclude the women to lessen their
workload at the individual questionnaire.
The sleeping away exclusion of the women which is one of the important points for
the completing the individual information is also evaluated. To estimate the level of
exclusion, the difference between the number of overnight visitors and the usual
residents sleep away is calculated. The three datasets seems have negligible problem
on the exclusion of women by the reason sleeping away.
The eligibility process of the women does not finish at the household questionnaire.
During the trainings of TDHS, the interviewers are well informed about the eligible
ages and warned about a situation where the woman may be found eligible at the
household list but may be found as out of eligible ages at individual questionnaire.
The interviewer facing such kind of problem is educated to turn back to individual
questionnaire and make the corrections and cancel the individual questionnaire
started to filled.
The assessment of the age data in individual questionnaire is studied at this study
with similar methods of evaluation used for the age data at household questionnaire.
In addition the completeness of the information is evaluated by the selected
background characteristics of woman. Myers index and the percent distribution of
the ages of 20-49 women are estimated for the assessment of digit preference and
heaping problem. The level of heaping at individual data is significantly high as
compared to the household data. The information collected directly from the
individual itself has low level of heaping than the data collected by a proxy
informant. The digits “0”, “3”, “5” and “8” are the preferred ones like in household
188
data. It is seen that the digit preference seems decreased over the years at TDHSs.
The structure of the age information at TDHS-1998 for different demographic
characteristics of women seems exceptional. Data gives fluctuation nearly at all
means of background characteristics. The results of TDHS-1993 and TDHS-2003
supports the previous studies implying the education of the women have positive
effect on the decrement of the level of heaping.
The completeness of the birth date data at individual questionnaire is also studied
with evaluating the level of imputation. The fertility rates will fluctuate with
regarding to the level of imputation. The overall level of imputation is very limited
for all three TDHSs. At TDHS-1998 the completeness of the age information is
lowest among three surveys. Approximately 10 % lower completeness is seen at
TDHS-1998 as compared to TDHS-1993 and TDHS-2003. The data collected from
urban is estimated as more complete than rural. The age of the woman on the other
hand seems related with the quality of the data. The level of imputation at the data
collected from women aged 30 or more is higher. Education seems have positive
effect on the completeness of the data. Both the education of the women and the
completeness level of the birth data increases interrelated. Women having secondary
or more education gave nearly complete birth date data.
The fertility and mortality rates at TDHS are calculated from the variables from the
ever married women questionnaire. Age of the mother and the birth and death date of
the children with relation to the interview date are used for the estimations. At that
point the quality of the birth data is studied in terms of birth and death date of the
children. The birth history section of the DHS questionnaire gives valuable
information on all the live births of the mother. At this study the quality of birth and
death information at birth history section is evaluated in terms of heaping,
displacement and completeness.
In general, the quality of the data at birth history at all three TDHSs seems well.
Especially for the last five years preceding the surveys almost complete birth date
information is collected from the field. During the training interviewers are well
189
informed to get complete birth and death month and year information with birth and
death age for five years preceding the survey.
The completeness of the birth history data seems not problematic in all three surveys.
However, among three TDHSs, more problems are seen at the TDHS-1998 data set
in terms of completeness of the data. Only 88.5 % of the women aged 45-49 gives
complete information for the births happened within 5 years prior to survey date. On
the contrary, results indicate that TDHS-2003 results give high level of complete
information for both birth and death dates of the children. In addition, the urban data
is seems better than the rural in terms of birth date. The common situation for the
cases at which the data is not complete, the year and age is complete and the month
is imputed. It is also seen that if the period where birth date of the children and the
interview date is increasing the remembrance problem will affect the data quality.
The education of women has positive effect on the completeness of the birth date
information. Women having secondary or more education gave almost complete
information for all age group of children. On the other hand, the sex of the child
seems not matter on the completeness of the information.
The Section 4 and 5 of the ever married woman questionnaire consists various and
detailed questions on children born within 5 years before the survey. The previous
studies on DHSs data quality argue that, the interviewers may displace the children
out of this eligible time period to escape from asking question at section 4 and 5
(Arnold, 1990). It is seen that although the levels are not high the displacement of
children to age 6 is quite common at TDHS. It is higher at TDHS-1998 and at urban
areas for all three surveys. As the displacement is a problem sourced mainly by the
interviewer, no relation between the characteristics of women is related.
The possible heaping problem at the birth data is also discussed. The displacement
problem leads to heaping of age 6 which is common for all level of estimations. For
the birth data although it is not very high heaping on “3”and “8” is also vivid.
Calculating the age of children with a heaped year of birth ending with “0” and “5”
190
leads to a heaping on “3” and “8”. As first the month and year of the birth is
collected than the age, the heaping is created by the years ending with “0” and “5”.
The incomplete birth date information may be resulted in the miscalculation of year
of birth for child. At the situation where the age of the child is known but the year
and month of the child is unknown, interviewer may just subtract the age from the
year of interview and calculate the year of birth. It is true if the month of birth is
earlier than the month of interview. However, if the month of interview is earlier
than the month of birth, the year of birth will be overestimated for one year. At this
study the difference between the imputed and the not imputed cases at the data entry
is assessed to understand the level of miscalculation of year of birth at birth history
data. It is clearly seen that, the imputed cases are more exposed to miscalculation of
year of birth problem than the not imputed ones. As mentioned before the
completeness of the all three surveys especially for the last 5 years is almost
complete. Hence, the imputed data which is mostly out of ages where fertility and
mortality rates are calculated from which have more possibility of miscalculation of
year will have no effect on indicators calculated from TDHS.
The problem of miscalculation of year of birth by imputing the month of the birth
before the survey month is almost same level at TDHS-1993 and TDHS-1998. These
two surveys were carried out at the same months of the year. However, TDHS-2003
fieldwork started at the end of 2003 and continued to the mid 2004 with some breaks
especially on winter months. Therefore the method to estimate the problem of
imputation of months did not worked well for TDHS-2003. The results indicate that
although the problem of imputing the birth month after the survey month is more or
less same at the first two surveys, the number of imputed cases is remarkably high at
TDHS-1998. Although this will create a problem if the number of cases where the
month of the birth is misplaced, as the imputed cases are nearly at the ages 6 or more
no effect will be on TFR or mortality rates.
The previous studies indicated that the coverage problem of the dead children is one
of the data quality problems in DHS especially at undeveloped or developing
191
countries. The average number of children ever born by age of mother in groups is
evaluated to catch the problem of coverage of live births. Results indicate that no
coverage problem is seen for all TDHSs except a possibility at TDHS-2003 at the
age group of 45-49 of mother.
The quality of the death data is also considered at this study starting with the
completeness of the date of birth data of the dead children with the comparison of
survivors. The completeness of the birth information of the dead children is lower
than the living children at all TDHSs. The urban data is in better conditions as
compared to the rural at first and the last TDHS. As the exceptional results are seen
at TDHS-1998; the rural data is more complete than the urban at this survey. The
education of the mother has positive effect of the completeness of the birth date
regardless of survival status of the children. It is noticeable that women having
secondary or higher education gave complete birth date information as compared to
women having lower education.
The defect at the death date data is mainly caused by the inconsistency of the
information with other data given by the mother. Although the date of death data for
the last 5 years preceding the survey comes complete from the field, results indicate
that for some cases the inconsistency of the information is seen as an for these cases
the imputation is done during the data entry. The inconsistent data is mainly a
problem of the data collected from the women having no education at all or did not
complete the primary level. It is seen that the interviewers must be alert for the
inconsistent dates at the birth history and if there is a problem of consistency it is
better to solve the problem during the interview with the women herself. A good
interviewer is not the person who asks the question as it is at the questionnaire and
record the answer suitable for the filling instructions. Interviewer must be careful
about the inconsistent answers of the women during the interview. The questions
asked to mother are not always the matter of mother in her daily life. If the women
have nothing to do with the birth date of the children and did not need to keep this
information in her mind, it is possible for her to give either inconsistent or no
information at all.
192
One of the common problems which will be effective on IMR and CMR is the
heaping on month 12 at the age at data. The recording instructions of DHS wants the
interviewer to record the death age in days if the answer is to one month, in months if
the answer is lower than 2 year and in years if it is 2 year or more. Previous studies
on DHS mention that a heaping on month 12 is expected because of the frequency of
the answer of “1 year”. Rounding the age at death information of the children to one
year results in the heaping at 12 months at the dataset is a common problem. During
the training, interviewers were told to probe the answer 1 year for age at death,
whether it is exactly 12 months or not. The extent of the problem is evaluated with
socio-demographic characteristic of the respondent. Although the general level of
heaping seems negligible among three TDHSs, TDHS-1998 has the highest heaping
among them. In addition, for all three surveys, the rural data is estimated as being
more vulnerable to the heaping compared with urban. The positive effect of the
higher education on the quality of age at death data in terms of heaping on month 12
is also seen. While the level of education of mother increases, heaping on month 12
decreases clearly.
One of the aims of the study is to evaluate the impact of the data quality on fertility
and mortality rates. By constructing simulations the effect of displacement of women
to non eligible ages and the women did not interviewed because of the sleeping away
factor on TFR and U5MR is assessed. Three simulations to evaluate the impact of
boundary effect exclusion and two simulations based on the exclusion of women by
sleeping away factor is applied to fertility data for all three surveys. Different fertility
patterns for the excluded women are taken into considers at these simulations seeing
the change at TFR. The overall results show that the estimated problems at the
household data in terms of exclusion of woman at eligible ages seems either no or
negligible effect on TFR estimated for Turkey. As, lower boundary effect was
calculated as insignificant in all three TDHS data, the effect of it on TFR is
ignorable. On the other hand, upper boundary effect estimated to be effective on East
region where the fertility of the women at last eligible group still continues. The
simulation where the excluded women at the upper boundary are supposed to have
193
has twice ASFR than the women at the same age group resulted in remarkably high
TFR at this region.
Three simulations are developed and applied to mortality data to evaluate the effects
of sleeping away and lower and upper boundary effects on under-five mortality rate.
Results indicate that simulation results based on upper boundary exclusion have no
effect on U5MR. Likewise, minor increase on U5MR is seen with the simulations of
lower boundary and sleeping away exclusion.
The impact of the heaping on month 12 at the age at death data on IMR and CMR is
also studied. The excess deaths which are calculated for each survey on month 12 are
redistributed to the infancy period and mortality rates are recalculated. Overall
results indicate that unimportant changes are seen at IMR. The change at CMR on
the other hand is also not in high levels. In general, the change at CMR is below 10
% for all surveys.
In general, the problems at the TDHS data seem negligible and have ignorable effect
on fertility and mortality rates. Among all the data quality of individual data at
TDHS-1998 seems problematic as compared to TDHS-1993 and TDHS-2003 in
terms of the variables evaluated at this study. The number of questions at Section 4
and 5 is highest at this survey. Although the overall level of displacement is at
acceptable levels, it is highest at this survey, the interviewers might exclude eligible
children for sections 4 and 5 to avoid the workload at these sections. Despite the
voluminous sections for the children born 5 years prior to survey, the average time to
complete the individual questionnaire is lowest. It seems that interviewers filled the
questionnaire in speed and the inconsistencies at the questionnaire are overlooked.
The probing which is the most effective tool increasing the data quality seems not
used at desirable levels at this survey.
The effect of long and divided fieldwork of TDHS-2003 will be effective on data
quality. Two groups of interviewers were selected and educated at this survey at
different dates. Although a standard education program is used, nearly 1.5 times
194
more interviewers were used at the field. This will lead to the increase the problems
caused by the inexperience at the field. It is clear that although you give good quality
training, the actual field experience adds knowledge and practice to the field staff.
Although the quality of the training of the interviewers is important, it does not solve
the problems at the quality of the data. The problems in the field will have an impact
on the work of interviewers. The important thing seems to give the training which
will initial keep the enthusiasm of the staff high. The field staff feeling her/him as a
part of the survey will behave different than the one aims just finishing “the job” as
soon as possible. The value of probing should be well taught to the interviewer.
The interviewers should be controlled either by the field editors or the supervisors at
the field either entering the interview with them or rechecking the basic information
at the questionnaires by revisiting the households. During the training just to mention
such kind of a control to the interviewers, they will keep the possibility of controlling
for any of the interview in their mind.
Field check tables should be used effectively to improve the data quality. The
interviewers who seems displacing the children or women to not eligible ages should
be warned. The result codes for the household and individual questionnaires should
be analyzed during the fieldwork. The teams where the response rates should be
tracked and alternative methods to increase the response rates should be used.
Revisiting the households where the household or individual questionnaire is not
completed at some time later either by the same team at the field or constructing a
new team will increase the response rates.
One of the alternative ways to increase the data quality seems using the Personnel
Data Assistant (PDA) at the field instead of printed questionnaires. The data entry
programs placed at PDA will give chance to interviewer to fill the questionnaire in a
digital format which makes the skips automatically and controls the inconsistencies
among the other responses. However, the system is new and the questions are
coming one by one to the small screen of the PDA. This will be an obstacle for the
195
individual to see the questionnaires as a whole. The flow of the questionnaire will be
disrupted by seeing the only one questionnaire at the screen. Another problem of
using only PDA’s at the field instead of printed questionnaires is the backup. As
there is no paper questionnaires are filled, a problem on the software or hardware
will result on the loss of data. Using PDAs or laptops for entering the questionnaires
with the software prepared specifically on data entry purpose will help the field staff
to catch the inconsistencies in a very short time so that interviewer may turn back to
the dwelling and correct the mistakes or inconsistencies.
At this study no external data is used to evaluate the differences at the TDHS data
and the registration system or other national studies. If the further studies on the
quality of the data of TDHS include such a comparison, this will give chance to
researchers make evaluation of the TDHS and external sources. The common and
alike problems should be studied within the data sets.
The effect of displacement of children to age 6 to escape the workload at section 4
and 5 at ever married woman questionnaire on fertility or mortality rates is not the
subject of this study. As the TFR is estimated for 3years preceding the survey the
displacement of the child to age above 5 will have no effect on these indicators. The
surveys where the TFR is estimated for 5 year period the impact of the displacement
can be studied but not in TDHSs. The further studies should include an evaluation on
the displacement of the age of children to age 6 or more on the TFR estimated for
five year period.
The neonatal and post-neonatal mortality rates are also estimated at TDHSs. The
ministry of health uses all mortality indicators for developing policy and plans. At
this study the effect of data quality of age at death data on IMR and CMR is studied.
The effect of data quality of age at death should be studied focusing on the neonatal
and post-neonatal infant mortality rates.
Basic Data Quality tables are presented at Appendix section of each TDHS. Standard
tables are published in a compact way. The only TDHS where the results are
196
discussed is TDHS-2003. In addition to the tables, a brief explanation of the tables
and a short discussion is made. It is recommended that this section of the report
should be more detailed and strengthen by making comparisons with the previous
surveys.
197
VII. REFERENCES
Al Abdel, M. 1987. Evaluation of Age Reporting in the 1983 Turkish Contraceptive
Prevalence and Family Health Status Survey. unpublished master thesis, HUIPS,
Technical Demography Department.
Albayrak, F. 1991. Yanlış Yaş Bildirimlerinin Türkiye Bazında Düzeltilmesi.
unpublished master thesis, HUIPS, Technical Demography Department.
Anderson, B. A., Katus, K., Puur , A. and Silver, B.D. 1994. “The Validity of Survey
Responses on Abortion: Evidence from Estonia” Demography, Vol. 31; No.1, 115-
132.
Araujo, E. and Sommer, L. (2002) “A summary history of eugenic theories and
practices in the United States” in Domain Errors: Cyberfeminist Practices!, Maria
Fernandez, Faith Wilding, Michelle M. Wright, eds , New York.
Arriaga, Eduardo E., and Associates. 1994. Population Analysis with
Microcomputers. Vol. 1: Presentation of Techniques. Washington, D.C.: U.S. Bureau
of the Census.
Arnold, F. 1990. Assesment of the Quality of Birth History Data in the Demographic
and Health Surveys. In An Assessment of DHS-I Data Quality, Methodological
Reports 1, Institute for Resource Development/Macro Systems, Inc. Columbia,
Maryland USA, pp. 81-111.
Bailey, M. and Makannah, T.J. 1996. “An Evaluation of Age and Sex Data of the
Population Censuses of Sierra Leone: 1963-1985” GENUS. Jan-Jun; 52(1-2):191-9.
198
Becker, S. and Sosa, D. 1992. “An Experiment Using a Month-by-Month Calendar in
a Family Planning Survey in Costa Rica” Studies in Family Planning, Vol. 23, No. 6.
(Nov. - Dec., 1992), pp. 386-391.
Becker, S.; Waheeb, Y.; El-Deeb, B.; Khallaf, N.; Black, R. 1996. “Estimating the
Completeness of Under-5 Death Registration in Egypt” Demography, Vol. 33, No. 3.
pp. 329-339.
Beckett, M.; Vanzo, J.D.; Sastry, N.; Panis, C.; Peterson, C. 2001. “The Quality of
Retrospective Data: An Examination of Long-Term Recall in a Developing Country”
The Journal of Human Resources, Vol. 36, No. 3. pp. 593-625.
Bixby, L.R., Céspedes J.H., Montero, D.A. and Seligson, M.A. 2005. “Improving
the Quality and Lowering Costs of Household Survey Data Using Personal Digital
Assistants (PDAs). An Application for Costa Rica” Centro Centroamericano de
Población of the Universidad de Costa Rica, 2005 meeting of the Population
Association of America. Philadelphia, March 31 to April 2, 2005.
Blake, J. 1983. “Book Review: E. Grebenik. World Fertility Survey Conference
1980: Record of Proceedings. 3 Vols. Voorburg, Netherlands, 1981. Vol. 1: 539 p.;
Vol. 2: 769 p.; Vol. 3: 574. P.” Population and Development Review, Vol. 9, No. 1,
pp. 153-156
Brass, W. 1996. “Demographic Data Analysis in Less Developed Countries: 1946-
1996” Population Studies, Volume 50, Number 3, November 1996, pp. 451-467(17)
Bowley, A.L. (1913) “Working-Class Households in Reading” Journal of the Royal
Statistical Society, Vol. 76, No. 7, pp. 672-701.
199
Ewbank, D.C. 1981. Age Misreporting and Age-Selective Underenumeration:
Sources, Patterns and Consequences for Demographic Analysis. Committee on
Population and Demography, Report No. 4. Washington, D.C.: National Academy
Press.
Canpolat, Ş. 2002. “Seçilmiş 10 İlde Yaş Bildirim Kalitesinin Değerlendirilmesi:
1990-2000 Genel Nüfus Sayımları ve 1997 Genel Nüfus Tespiti”, unpublished.
Statistics Days 2002. Hacettepe University, Department of Statistics, 16 – 18 May
2002.
Canpolat, Ş. 2003. Türkiye’de Gerçekleştirilen Nüfus Sayımlarinda Yaş Bildirim
Kalitesinin Analizi. unpublished Expertness Thesis. State Institute of Statistics,
Ankara.
Chidambaram, V.C. and Pullum T.W. (1981) “Estimating Fertility Trends from
Retrospective Birth Histories: Sensitivity to Imputation of Missing Dates”
Population Studies, Vol. 35, No. 2. pp. 307-320
Chidambaram, V.C., John G. Cleland, and Vijay Verma. 1980. Some aspects of WFS
Data Quality: A Preliminary Assessment. WFS Comparative Studies, No. 16.
Voorburg, Netherlands: International Statistical Institute.
Chidambaram, V.C. and Zeba A. Sathar. 1984. Age and Date Reporting. WFS
Comparative Studies, No.5. Voorburg, Netherlands: International Statistical Institute.
Curtis, S.L. 1995. Assesment of the Quality of the Data Used for the Direct
Estimation of Infant and Child Mortality in the DHS-II Surveys. Macro International
Inc. Maryland.
Curtis, S.L. and Arnold, F. 1994. An Evaluation of the Pakistan DHS Survey Based
on the Reinterview Survey. Occasional Papers 1. Macro International Inc. Calverton,
Maryland.
200
Charles W. W. 1985. An Analysis of the Quality of Data in the 1982 Puerto Rico
Fertility and Family Planning Assessment. pp.355-357.
Cleland , J. 1986. “Fertility and Family Planning Surveys: Future Priorities in the
Light of Past Experiences” International Family Planning Perspectives, Vol. 12, No.
1. pp. 2-7.
Croft, T. 2004. DHS Data Editing and Imputation. www.macrodhs.com (04.12.2004)
Çavdar, T. ,F. Karadayı, H. Serinken, K.S. Srikantan, S. Timur, 1971. Türkiye’de
Aile Yapısı ve Nüfus Sorunları Araştırmasının Veri Toplama Teknikleri (1968).
Hacettepe Üniversitesi Yayınları: D-9. Ankara.
Das Gupta, P. 1975. “A General Method of Correction for Age Misreporting in
Census Populations” Demography, Vol. 12, No. 2 (May, 1975), pp. 303-312
Demeny, P. and Shorter, F. 1968. Estimating Turkish Mortality, Fertility and Age
Structure. Statistics Institute, Istanbul University.
Feeney, G. 1980. Estimating Infant Mortality Trends from Child Survivorship Data.
Population Studies, 34 (1), March, pp. 102-128
Feeney, G. 1991. Child Survivorship Estimation: Methods and Data Analysis. Asian
and Pacific Population Forum. Vol. 5, Nos. 2-3, Summer –Fall.
Goldman, Noreen, Shea O. Rutstein, and Susheela Singh. 1985. Assessment of the
Quality of Data in 41 WFS Surveys: A Comparative Approach. WFS Comparative
Studies, No. 44. Voorburg, Netherlands: International Statistical Institute.
Güneş, M. 1989, Nüfus Sayımlarında Hatalı Yaş Beyanlarının Düzeltilme
Yöntemleri, Devlet İstatistik Enstitüsü Eğitim Merkezi Seminer Çalışması, Ankara.
201
Hacettepe Üniversitesi (1978) Türkiye’de Nüfus Yapısı ve Nüfus Sorunları - 1973
Araştırması. Hacettepe Üniversitesi Yayınları, Ankara.
Hacettepe University Institute of Population Studies (HUIPS) (1980) 1978 Turkish
Fertility Survey. Hacettepe University, Institute of Population Studies, Ankara.
Hacettepe University Institute of Population Studies (HUIPS) (1987) 1983 Turkish
Population and Health Survey. Hacettepe University Institute of Population Studies,
Ankara.
Hacettepe University Institute of Population Studies (HUIPS) (1989) 1988 Turkish
Population and Health Survey. Hacettepe University Institute of Population Studies,
Ankara.
Hacettepe University Institute of Population Studies (HUIPS) and Macro
International Inc. 1999. Turkey Demographic and Health Survey 1998, Ankara.
Hacettepe University Institute of Population Studies (HUIPS) 2004. Turkey
Demographic and Health Survey, 2003, Hacettepe University Institute of Population
Studies, Ministry of Health General Drectorate of Mother and Child Health and
Family Planning, State Planning Organization and European Union, Ankara, Turkey.
Hacettepe University Institute of Population Studies (HUIPS) 2008. Surveys
http://www.hips.hacettepe.edu.tr/english/surveys.htm date on: 14.04.2008
Hancıoğlu, A. 1997. “Fertility Trends in Turkey: 1978-1993”, Fertility Trends,
Women’s Status, and Reproductive Expectations in Turkey. HUIPS and Macro
International Inc. Calverton, Maryland.
Hobcraft, J. N.; Goldman, N.; Chidambaram, V. C. 1982. “Advances in the P/F Ratio
Method for the Analysis of Birth Histories” Population Studies, Vol. 36, No. 2.
pp. 291-316.
202
Institute for Resource Development/Macro Systems, Inc. 1990. An Assessment of
DHS-I Data Quality, Methodological Reports 1, Columbia, Maryland USA, pp. 113-
137.
Ishak, M. 1999. “Regression methods to model incomplete population data” Bulletin
of the International Statistical Institute, 52nd Session, Proceedings,Tome LVIII,
Finland.
Kalton, G. (1998) SIPP Quality Profile. Survey of Income and Program
Participation, SIPP Working Paper Number 230, 3rd Edition. U.S. Department of
Commerce, Bureau of the Census.
Killion, R. A. (2004) Survey of Income and Program Participation (SIPP) 2004
Panel: Source and Accuracy Statement for Wave 1 - Wave 7 (core) Public Use Files
(S&A-7). U.S. Department of Commerce, Bureau of the Census.
Kingkade, W.W.; Sawyer, C.C (2001) “Infant Mortality in Eastern Europe and the
Former Soviet Union Before and After the Breakup” Prepared for presentation at the
2001 Meetings of the International Union for the Scientific Study of Population,
Salvador de Bahia, Brazil, August 19-24.
Knodel, J.; Piampiti, S. 1977. “Response Reliability in a Longitudinal Survey in
Thailand” Studies in Family Planning, Vol. 8, No. 3. pp. 55-66.
Koç, İ. 2004. “Appendix- D. Data Quality”. Turkey Demographic and Health Survey,
2003, Hacettepe University Institute of Population Studies, Ministry of Health
General Directorate of Mother and Child Health and Family Planning, State Planning
Organization and European Union, Ankara, Turkey.
Loaiza, E. 2004. Does data quality explain the differences in the current global
estimates for mortality and education? Committee for the Coordination of Statistical
203
Activities, Conference on Data Quality for International Organizations, Wiesbaden,
Germany, 27 and 28 May 2004.
Marckwardt, A.M.; Rutstein S.O. 1996. Accuracy of DHS-II Demographic Data:
Gains and Lose in Comparison with Earlier Surveys. DHS Working Papers Number
19. Macro International Inc. Calverton, Maryland, USA.
Macro International Inc. 2008. DHS Overview
http://www.measuredhs.com/aboutsurveys/dhs/start.cfm date on: 04.07.2008.
Mauldin WP, Watson WB. and Noe LF.(1970) “KAP surveys and evaluation of
family planning programs” Proceedings of the Ford Foundation Conference on
Evaluation, Marino, Italy, April 20-24.
Ministry of Health (Turkey) (MOH), Hacettepe University Institute of Population
Studies and Macro International Inc. 1994. Turkey Demographic and Health Survey
1993, Ankara, Turkey.
Mohammed, M. A. and Iqbal H. Shah. 2000. Sanctions and Childhood Mortality in
Iraq. The Lancet, Vol 355, May 27. pp. 1851-1957.
Mukherjee B.N. and Mukhopadhyay B.K. 1988. “A Study of Digit Preference and
Quality of Age Data in Turkish Censuses” GENUS. 1988 Jan-Jun;44(1-2):201-27.
Nadezhda A. and Redmond,G. 2003. How High is Infant Mortality in Central and
Eastern Europe and the CIS? Innocenti Working Papers No. 95. UNICEF Innocenti
Research Centre, Florence, Italy.
Nagi, Moni H., E.G. Stockwell, and L.M. Snavley. 1973. Digit Preference and
Avoidance in the Age Statistics of Some Recent African Censuses: Some Patterns
and Correlates. International Statistical Review. 41 (2): 165-174.
204
Ntozi J.P. 1978. “The Demeny-Shorter and Three-Census Methods for Correcting
Age Data” Demography, Vol. 15, No. 4 (Nov., 1978), pp. 509-521
Poston, D.L.Jr., Iris Hye Jin Chu, Jaime M. Ginn, Godfrey Jin-Kai Li, Catherine
Hong Vo, Carol S. Walther, Ping Wang, Julie Juan Wu, and Michael Ming Yuan.
2000. “An Analysis of the Quality of the Age and Sex Data of the Republic of Korea
and Its Provinces, 1970 and 1995.” pp. 3-47 in Han Gon Kim (editor), International
Conference on Projections and Policy Implications for the Issues of the Elderly in
the 21st Century. Taegu, South Korea: Institute of Gerontology, Yeungnam
University.
Potter, Joseph E. 1997. Problems in Using Birth History Analysis to Estimate Trends
in Fertility. Population Studies. 31(2): 335-364.
Preston, S. H., Elo, I.T. and Stewart, Q.. 1998. Effects of Age Misreporting on
Mortality Estimates at Older Ages. PARC Working Paper Series, University of
Pennsylvania, Population Aging Research Center.
Pullum, T.W.; Harpham, T. and Ozsever, N. 1986. “The Machine Editing of Large-
Sample Surveys: The Experience of the World Fertility Survey” International
Statistical Review / Revue Internationale de Statistique, Vol. 54, No. 3. pp. 311-326.
Pullum, T. 1991. “Statistical Methods to Adjust for Date and Age Misreporting to
Improve Estimates of Vital Rates in Pakistan” Statistics in Medicine. Vol. 10’ pp.
191-200.
Pullum, T.W. 2006. An Assessment of Age and Date Reporting in the DHS Surveys,
1985-2003. Methodological Reports No. 5 Calverton, Maryland: Macro
International. Inc.
Retherford, R. D.; Mishra, V. K.; Prakasam, G. (2001) How Much Has Fertility
Declined in Uttar Pradesh? National Family Health Survey Subject Reports,
205
Number 17, East-West Center, Population and Health Studies, Honolulu, Hawaii,
U.S.A.
Retherford, R. D.; Mishra, V. K. (2001) An Evaluation of Recent Estimates of
Fertility Trends in India. National Family Health Survey Subject Reports, Number
19, East-West Center, Population and Health Studies, Honolulu, Hawaii, U.S.A.
Rindfuss, Bumpass, L.L. and Palmore, J.A. 1987. “Analyzing Fertility Histories: Do
Restrictions Bias Results?”. Demography 24(1). pp. 113-122.
Rutstein and Bicego, 1990 “Assesment of the Quality of Data Used to Ascertain
Eligibility and Age in the Demographic and Health Surveys” In An Assessment of
DHS-I Data Quality, Methodological Reports 1, Institute for Resource
Development/Macro Systems, Inc. Columbia, Maryland USA, pp. 113-137.
Rutstein, S.O. and Rojas, G. 2003. Guide to DHS Statistics. Demographic and Health
Surveys, ORC Macro, Calverton, Maryland, USA.
Scott, Christopher and G. Sabagh. 1970. The Historical Calendar as a Method of
Estimating Age: The Experience of the Moroccan Multi- Purpose Survey of 1961-63.
Population Studies, 24(1): 93-109
Shryock, Henry S., Jacob S. Siegel, and Associates. 1976. The Methods and
Materials of Demography. Condensed Edition by Edward G. Stockwell. New York:
Academic Press.
Shryock, H.S. and Taeuber, C. 1976. The Conventional Population Census.
Labaratories for Population Statistics, Scientific Report Series No. 25.
Stockwell, E.G. 1966. Patterns of Digit Preference and Avoidance in the Age
Statistics of Some Recent Nation Censuses: A Test of the Turner Hypothesis.
Eugenics Quaterly. 13(3):205-208
206
Sullivan, J.M., Bicego, G. and Rutstein. S.O. 1990. “Assesment of the Quality of
Data Used for the Direct Estimation of Infant and Child Mortality in the
Demographic and Health Surveys”. In An Assessment of DHS-I Data Quality,
Methodological Reports 1, Institute for Resource Development/Macro Systems, Inc.
Columbia, Maryland USA, pp. 113-137.
Timur, S. (1972) Türkiye’de Aile Yapısı. Hacettepe University, D-15, Ankara.
Törüner, K. 2001. Comparability of Data Collected in Turkish Demographic
Surveys, unpublished master thesis, HUIPS, Economic and Social Demography
Department.
Tungul, B. 1995. Türkiye’de Yapılan Nüfus Sayımlarında Kapsam ve Cevaplama
Hataları. Unpublished master thesis, HUIPS Technical Demography Department.
Turkish Statistical Institute (TURKSTAT), 2008. “Genel Nüfus Sayımı”
http://www.die.gov.tr/nüfus_sayimi.htm date on: 03.02.2008
Turner, Stanley H. 1958. Patterns of Heaping in Reporting of Numerical Data.
Proceedings of the Social Statistics Section. Washington, D.C.: American Statistical
Association.
United Nations (UN), 1967. Manual II. Methods of Appraisal of Quality of Basic
Data for Population Estimates. Population Studies, No. 23. New York: Dept. Of
Economic and Social Affairs
United Nations (UN), 2004. Handbook on the Collection of Fertility and Mortality
Data. Department of Economic and Social Affairs Statistics Division. Studies in
Methods Series F No.92.
207
United Nations (UN), 1987. A Comperative Evaluation of Data Quality in Thirty-
Eight World Fertility Surveys. New York: United Nations, Department of
International Economic and Social Affairs.
United Nations Economic Commission for Europe (UNECE), 2008.
http://www.unece.org/pau/ggp/keyfeatures.htm date on: 05.05.2008
Üner, S. 1983. Evaluation of the Turkish Fertility Survey, WFS Scientific Report
No. 43, London: World Fertility Survey, International Statistical Institute,
Netherlands.
Vaessen, M. 2008. “Demographic Surveys, History and Methodology of”
http://www.novelguide.com/a/discover/epop_01/epop_01_00086.html, date on:
03.02.2008.
Warren, C.W. 1985.An Analysis of the Quality of Data in the 1982 Puerto Rico
Fertility and Family Planning Assessment, Proceedings of the Survey Research
Methods Section, American Statistical Association.
Yavuz S. and Coşkun Y., “1935’den Günümüze Nüfus Sayımları ve 1997 Nüfus
Tespitindeki Yaş ve Cinsiyet Verisinin Değerlendirilmesi” unpublished. Statistics
Days 2002. Hacettepe University, Department of Statistics, 16 – 18 May 2002.
208
VIII. ANNEXES
209
ANNEX VIII.1.
Table VIII.1. Nationwide Demographic Surveys of Turkey
Year and the name of the Survey Instituti
on
Size and Frame Questionnaire Types
1963 Turkish Demographic Survey SPH 9,701 households,
~8000 eligible women
- household
- ever married women
1965-1968 Turkish Demographic Survey SPH 240,000 households - household
1968 Turkish Population Structure and Population
Problems Survey
HUIPS 4505 househods
3303 eligible women
2787 eligible husbands
- household
- married women
- husband
- general information of village and small
town
1973 Turkish Population Structure and Population
Problems Survey
HUIPS 6500 households
4580 eligible women
- household
- currently married women
- divorced/widow women
- people employed abroad
- general information of village and small
town
1974-1975 Turkish Population Survey SIS 17327 households - household
- individual
1978 Turkish Fertility Survey HUIPS 5137 households
4769 eligible women
- household
- individual
1983 Turkish Fertility and Health Survey HUIPS 6545 households
5398 eligible women
- household
- individual
1988 Turkish Population and Health Survey HUIPS 6552 households
5257 eligible women
2264 eligible husbands
- household
- women
- husband
1989 Turkish Demographic Survey SIS 17675 Households - household
1993 Turkey Demographic and Health Survey HUIPS 8619 households
6519 eligible women
- household
- women
1998 Turkey Demographic and Health Survey HUIPS 8059 households
8576 eligible women
1971 eligible husbands
- household
- ever married women
- never-married women
- husband
2003 Turkey Demographic and Health Survey HUIPS 10816 households
8075 eligible women
- household
- ever married women
210
ANNEX VIII.2.
Table VIII.2.1. Myers, Bachi and Whipple Indices for Total Population at
Household Data by Region and Type of Place of Residence, TDHS 1993, 1998
and 2003
Myers Index
(Total)
Bachi Index
(Total)
Whipple Index
(Total)
Region
West
1993 12.9 8.1 1.23
1998 6.3 3.2 1.03
2003 6.6 4.1 1.07
South
1993 12.5 9.0 1.28
1998 11.4 6.8 1.21
2003 6.2 5.1 1.09
Central
1993 16.1 11.1 1.33
1998 10.4 6.3 1.18
2003 8.0 5.3 1.12
North
1993 17.6 11.2 1.41
1998 9.9 6.3 1.21
2003 8.6 5.6 1.05
East
1993 25.8 20.5 1.94
1998 16.9 13.0 1.58
2003 10.5 7.8 1.31
Type of Place of Residence
Urban
1993 14.4 9.8 1.31
1998 8.0 4.5 1.10
2003 6.3 4.4 1.10
Rural
1993 18.3 13.4 1.53
1998 15.4 9.8 1.36
2003 9.0 6.6 1.16
Total
1993 16.2 10.9 1.39
1998 10.0 6.3 1.19
2003 7.4 5.1 1.12
211
Figure VIII.2.1. Bachi Preference by Digit, TDHS-1993
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
6.0
0 1 2 3 4 5 6 7 8 9
TURKEY: TDHS-1993 Total
Male Female
Bachi Preference by Digit
Figure VIII.2.2. Bachi Preference by Digit, TDHS-1998
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
0 1 2 3 4 5 6 7 8 9
TURKEY: TDHS-1998 Total
Male Female
Bachi Preference by Digit
212
Figure VIII.2.3. Bachi Preference by Digit, TDHS-2003
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
0 1 2 3 4 5 6 7 8 9
TURKEY: TDHS-2003 Total
Male Female
Bachi Preference by Digit
213
ANNEX VIII.3.
Table VIII.3.1. Percentage of Children Whose Month of Birth Falls in the
Month of Interview or Earlier by Demographic Characteristics of Women,
TDHS 1993, 1998 and 2003
Imputed Cases Number of
Children
Non-Imputed Cases Number of
Actual Expected Actual Expected Children
Age
15-19
1993 - - - 80.9 74.5 177
1998 - - - 80.1 72.2 159
2003 - - - 59.4 61.2 140
20-24
1993 84.4 77.9 14 77.8 73.1 1373
1998 94.9 71.9 24 77.0 72.4 1166
2003 25.7 46.9 9 51.1 52.8 1240
25-29
1993 73.8 73.2 30 78.5 73.7 2674
1998 90.5 74.0 118 77.3 72.8 2266
2003 52.6 54.3 76 53.9 54.3 2770
30-34
1993 92.4 75.2 73 78.6 73.4 3894
1998 85.1 73.4 306 76.1 72.7 2804
2003 42.1 49.7 169 55.2 54.2 3771
35-39
1993 84.6 73.2 130 79.4 73.6 4045
1998 85.8 72.7 501 76.9 72.4 3036
2003 39.5 46.8 257 56.1 55.2 4082
40-44
1993 82.8 72.2 226 80.7 73.3 3844
1998 90.2 73.4 707 75.9 73.1 2794
2003 45.9 50.9 401 57.4 56.5 4318
45-49
1993 90.6 72.6 210 79.5 73.5 3136
1998 88.1 71.9 823 79.2 73.2 2504
2003 36.8 45.4 438 56.9 55.1 3563
214
Table VIII.3.1. Percentage of Children Whose Month of Birth Falls in the
Month of Interview or Earlier by Demographic Characteristics of Women,
TDHS 1993, 1998 and 2003 (Continued)
Imputed Cases Number of
Children
Non-Imputed Cases Number of
Actual Expected Actual Expected Children
Education
No educ/Pri.Inc. 87.5 72.9 547 80.9 73.27 9495
1993 87.9 72.5 2007 77.4 72.1 5372
1998 39.0 46.5 1074 50.3 48.9 6362
2003
Primary
1993 80.7 73.5 134 77.7 73.7 8642
1998 88.4 73.8 464 76.3 73.1 8139
2003 50.6 55.3 273 57.6 57.1 11460
Secondary
1993 100.0 70.8 2 77.4 73.3 1007
1998 100.0 73.1 9 80.3 73.9 1218
2003 100.0 84.4 2 62.7 62.7 2003
Total
1993 86.2 72.9 683 19.3 73.5 19144
1998 88.1 72.8 2480 77.1 72.8 14729
2003 41.5 48.4 1349 55.8 55.0 19824
215
Table VIII.3.2. Percentage of Children Whose Month of Birth Falls in the
Month of Interview or Earlier by time period of interviewer in the field, TDHS
1993, 1998 and 2003
Imputed Cases Number of
Children
Non-Imputed Cases Number of
Actual Expected Actual Expected Children
Time period of interviewer in the field
1st week
1993 78.5 66.7 96 72.5 66.9 2409
1998 78.8 66.8 298 73.1 67.0 1646
2003 91.5 91.2 92 94.7 94.6 2521
2nd week
1993 79.3 66.7 110 71.3 66.7 2480
1998 82.8 66.8 282 70.5 66.9 1683
2003 81.9 82.7 125 94.3 94.3 2144
3rd week
1993 91.4 69.6 112 75.5 69.2 2884
1998 85.7 67.1 323 72.1 67.3 2009
2003 88.2 8826 161 93.4 92.8 2286
4th week
1993 93.8 75.0 82 82.0 75.0 2436
1998 85.8 68.5 359 74.3 69.2 1989
2003 83.3 86.2 131 96.7 96.7 2394
More
1993 87.2 78.3 284 83.8 78.1 8934
1998 92.8 78.4 1218 81.5 77.9 7402
2003 14.5 25.0 840 20.9 19.7 10478
Total
1993 86.2 72.9 683 19.3 73.5 19144
1998 88.1 72.8 2480 77.1 72.8 14729
2003 41.5 48.4 1349 55.8 55.0 19824
216
ANNEX VIII.4.
Figure VIII.4.1. Age at Death for Children died Less Than Two Years of Age,
Turkey, TDHS-1993
Figure VIII.4.2. Age at Death for Children died Less Than Two Years of Age by
Region, TDHS-1993
217
Figure VIII.4.3 Age at Death for Children died Less Than Two Years of Age by
Type of Place of Residence, TDHS-1993
Figure VIII.4.4. Age at Death for Children died Less Than Two Years of Age by
Education of Women, TDHS-1993
218
Figure VIII.4.5. Age at Death for Children died Less Than Two Years of Age,
Turkey, TDHS-1998
Figure VIII.4.6. Age at Death for Children died Less Than Two Years of Age by
Region, TDHS-1998
219
Figure VIII.4.7. Age at Death for Children died Less Than Two Years of Age by
Type of Place of Residence, TDHS-1998
Figure VIII.4.8. Age at Death for Children died Less Than Two Years of Age by
Education of Women, TDHS-1998
220
Figure VIII.4.3. Age at Death for Children died Less Than Two Years of Age,
Turkey, TDHS-2003
Figure VIII.4.10. Age at Death for Children died Less Than Two Years of Age
by Region, TDHS-2003
221
Figure VIII.4.11. Age at Death for Children died Less Than Two Years of Age
by Type of Place of Residence, TDHS-2003
Figure VIII.4.12. Age at Death for Children died Less Than Two Years of Age
by Education of Mother, TDHS-2003
222
ANNEX VIII.5.
Table VIII.5.1. Estimates of Infant and Child Mortality for the Five to Ten year
Period before the Survey, Adjusted for Heaping of Deaths at Twelve Months of
Age, by Region and Type of Place of Residence, TDHS 1993, 1998, 2003.
5-9 years Infant Mortality (1q0) Child Mortality (4q1)
Unadjusted Rate Adjusted
Rate
Percent
Increase
Unadjusted Rate Adjusted
Rate
Percent
Decrease
Region
West
1993 55.98 55.98 0.0 13.77 13.77 0.0
1998 51.83 52.81 1.9 11.74 10.73 8.6
2003 31.20 31.37 0.6 2.14 1.97 8.2
South
1993 73.87 75.68 2.4 16.98 15.11 11.0
1998 48.40 48.65 0.5 10.66 10.42 2.3
2003 41.47 41.90 1.0 16.69 16.27 2.5
Central
1993 84.43 85.14 0.8 18.70 17.97 3.9
1998 48.21 48.82 1.3 11.42 10.78 5.6
2003 35.58 35.58 0.0 8.48 8.48 0.0
North
1993 88.57 89.05 0.5 17.38 16.88 2.9
1998 63.57 68.22 7.3 32.02 31.55 1.5
2003 45.78 47.92 4.7 10.35 8.08 21.9
East
1993 106.20 106.84 0.6 18.72 18.02 3.7
1998 61.32 61.96 1.0 15.04 14.35 4.6
2003 76.88 77.16 0.4 17.68 17.38 1.7
Type of Place of Residence
Urban
1993 70.40 70.73 0.5 12.53 12.19 2.7
1998 49.05 49.79 1.5 11.58 10.81 6.6
2003 37.76 37.95 0.5 8.67 8.46 2.4
Rural
1993 96.17 97.23 1.1 22.78 21.66 4.9
1998 62.27 63.54 2.0 17.97 16.64 7.4
2003 63.72 64.30 0.9 12.24 11.66 4.8
Total
1993 83.71 84.37 0.8 17.01 16.33 4.0
1998 54.92 55.87 1.7 14.04 13.06 7.0
2003 47.23 47.56 0.7 9.91 9.58 3.4