APPENDIX A
Coverage, Definitions, Methods, and Data
This appendix describes operational definitions adopted by the panel, approaches to data analysis, and chapter by chapter discussions of the specific data choices made and the reasons for those choices. The first section of the appendix presents data and methodological issues pertaining to the entire report as well as an introduction to Demographic and Health Survey (DHS) data, which are the only data used extensively in more than one chapter. The second section describes the data and methods used in each of the chapters in Parts II and III of the report.
ENTIRE REPORT
Coverage
When data are presented in Chapters 3-8, countries are defined by the panel as developing if they fit the following criteria: (1) location in Africa, Asia, or Latin America and the Caribbean, using United Nations (UN) regional groupings, and (2) classification by the World Bank (2002b) as low, lower-middle, and upper-middle-income countries as of 2000. When the panel judged there to be sufficient coverage in terms of countries within regional groupings for a particular topic, regional averages were presented according to geographic and income categories. For the summary data on the size and distribution of young people in developing countries presented in Table 2-1, we use the United Nations definition of developing countries: “Less developed regions comprise all regions of Africa, Asia (except Japan),
Latin America and the Caribbean plus Melanesia, Micronesia and Polynesia” (United Nations, 2003d:46).
Regional Groups
Geographical Categories: The panel grouped developing countries into eight geographic regions, which were constructed from the geographic subregions used by the United Nations for its population estimates and projections (United Nations, 2003b). For the purposes of the panel report, Latin America and the Caribbean consists of two subregions (the Caribbean and Central America1 and South America), sub-Saharan Africa consists of two subregions (Western and Middle Africa and Eastern and Southern Africa), Asia includes three subregions (Eastern Asia, South-central, and Southeastern Asia, and former Soviet Asia, which includes eight former Soviet countries from South-central and Western Asia), and the Middle East combines two subregions (Western Asia and Northern Africa). Wherever the data coverage was deemed sufficient to allow population-weighted estimates by geographic region, these are the categories used. This was typically the case for the DHS survey data as well as for the UN data base on marriage prevalence.
Income Groups: The World Bank classifies countries (“economies”) into four economic groups. In 2000 the range for each of these four groups was as follows: low income (gross national income of $755 or less), lower-middle income ($756 to $2,995), upper-middle income ($2,976 to $9,266), and high income ($9,267 or more) (World Bank 2002b).2 Whenever the data allowed estimates by country income category, the panel used the first three country income categories described above.
1 |
Mexico is included in the Central America and Caribbean region even though in some international contexts (e.g., NAFTA) it is considered to be part of North America. |
2 |
Gross national income is in current U.S. dollars converted using the World Bank Atlas method. The purpose of the Atlas conversion factor is to reduce the impact of exchange rate fluctuations in the cross-country comparison of national incomes. A full explanation can be found on page 380 of World Bank (2002b). Updates of country income groupings on the World Bank Website: http://www.worldbank.org/data/countryclass/classgroups.htm as of September 30, 2002, led the panel to make a few adjustments to these country groupings, including shifting Korea into the high-income group (and therefore out of the developing country group) and adding East Timor in the low-income category, for which no data on income were provided in World Bank (2002b). |
Approaches to Data Analysis
Trends: The working group decided to concentrate primarily on data on recent trends, ideally the last two decades of the twentieth century wherever possible. In cases in which recent trend data were not available, we focused primarily on data that were sufficiently recent to characterize contemporary patterns and interrelationships.
Two approaches were used to measure trends: (1) a comparison of comparable data collected for surveys or censuses at two separate points in time, and (2) cross-cohort comparisons within a single recent survey based on the retrospective reporting of specific experiences or events during the transition to adulthood.
Age Ranges: The working group decided not to set a rigid range of ages but to explore broad age ranges that would allow a full exploration of transitions. For practical purposes, this usually meant concentrating on ages 10-24, but not always when transitions were found to continue until later ages. In several cases, constraints associated with available data limited our ability to analyze the most conceptually appropriate age group. For example, the DHS surveys restrict their samples of individual women or men, generally to the 15-49-year-old age range. The World Values Surveys do not interview youth younger than age 18. Mortality data by cause as estimated by the World Health Organization (WHO) are presented only for an aggregated age group: 15-29.
Population Weights: To provide the basis for statements about broad regional and global trends in cases in which there were sufficient data from each regional grouping to allow some representativeness, data were aggregated by region and income level and weighted according to the size of the population of young people ages 10-24 in 2000 (United Nations, 2001). The aggregated data were produced by calculating a weighted average of a particular statistic using as a weight for each country the percentage of the population of young people from countries with available data that are estimated to reside in that country. When data are available for two periods in time, we weighted both figures with the UN 2000 population estimates. Holding the weight constant allows us to attribute any apparent change between the two years to changes in the indicator rather than changes in the weights (allowing us to look at the average change over the time span).
Demographic and Health Surveys
The panel used Demographic and Health Surveys extensively for Chapter 3 on schooling, for a few topics in Chapter 4 (sex and contraceptive
use), extensively in Chapter 7 on marriage, and exclusively in Chapter 8 on parenthood. Although other data on these topics are available, the DHS data have the special advantage of allowing comparative analysis over time and across countries while controlling for important social and economic factors like urban and rural residence, socioeconomic status, and education level.
For each topic treated, all DHS data from publicly accessible surveys fielded since 1990 that included data for the topic in question were used (see Table A-1). The number of countries included in each analysis varied from 49 to 52 in the education, marriage, and parenthood chapters to as few as 39 in the case of a few topics—sex and contraceptive use—presented in the health chapter. This smaller sample of countries was necessitated because data on these topics for adolescents were not available for countries in which the DHS data were limited to ever-married women samples.
Using the full sample of 52 countries, we can make certain generalizations about the timeliness and representativeness of the DHS. Survey years range from 1990 to 2001 and all but 7 of the 52 surveys were conducted post-1995. The median year of the 52 surveys is 1998. Fewer surveys (32) include data on men, mostly but not exclusively in Latin America and the Caribbean and Africa. The specific number of surveys included in each analysis varies for a variety of reasons, including (1) whether the survey interviewed all women regardless of marital status or only ever-married women of reproductive age, (2) whether a specific question was asked, and (3) whether the data were available at the time the tables were compiled.
Table A-1 provides a listing of all the DHS data sets included in the panel sample along with sample sizes. The sample sizes for the survey of women of reproductive age varied from a little over 3,000 for the small island of Comoros to over 90,000 for the most recent sample in India. Typically, the samples of men are much smaller in size. The DHS data also include a household roster with some basic data on all household members, including the schooling level and current schooling participation of all household members.
Together, the 52 DHS surveys of females provide representative data for 90 percent of the population of low-income countries as defined by the World Bank in 2000, 19 percent of the population in lower middle-income countries, and 53 percent of upper middle-income countries (see Table A-2 for regional and income categories). Thus, overall the DHS data are most representative of the experience in the poorest countries. In addition, DHS surveys of females are representative of the populations in Eastern and Southern Africa (92 percent), South-central and South-eastern Asia (86 percent), Western and Middle Africa (75 percent), South America (74 percent), and the former Soviet Asia (68 percent), but they are less representative of all women in the Middle East (55 percent) and the Caribbean and
Central America (21 percent) (see Table A-2). They include no countries from Eastern Asia, most importantly China, which has over a fifth of the developing world’s young people.
DATA FOR INDIVIDUAL CHAPTERS
The panel’s working group on data reviewed data on each topic for quality and coverage before deciding whether or not to treat a topic and, when choices were available, determined which data were best to provide the most accurate and comprehensive information on a particular topic.
Chapter 3: Schooling
Since 1990, the Demographic and Health Surveys have included comparable questions on school attendance and attainment in all household surveys. In the absence of comparable census data, we have chosen to rely on DHS data for this report, supplemented when available by data on schooling from other household surveys to describe trends and patterns of schooling participation and attainment. These surveys allow us to calculate not only current schooling attendance rates, but also various cohort measures of schooling participation, grade attainment and progression between levels to estimate trends consistently over the past 20 years—a particular goal for the overall panel report. They also allow us to report current attendance figures according to household wealth categories and urban-rural residence. Before making this choice of data source as our primary source for the analysis of schooling patterns and trends, we carefully reviewed the pros and cons of United Nations Education, Scientific and Cultural Organization (UNESCO) data versus DHS data. These considerations are discussed below.
These DHS data are supplemented for Latin America with a collection of household surveys assembled by the Inter-American Bank from the mid-1980s to mid-1990s that covers roughly 93 percent of the population of Latin America and includes more countries than available from DHS in the upper middle-income category (Behrman, Duryea, and Székely, 1999a) and for China with data on eight provinces from the China Health and Nutrition Survey (Hannum and Liu, 2005).3
The enrollment ratios published by UNESCO are the most widely used international statistics on education for measuring progress over time and
TABLE A-1 List of Countries with Recent DHS Surveys Used in Analysis, Including Estimates of Youth Population, Dates of Surveys, Sample Sizes, and Chapters in Which Data Were Used
Country |
Survey Date(s) |
Youth Population, 10-24 |
|
UN Data Base 2000 (in thousands) |
Household Sample Size |
||
Armenia |
2000 |
1,068 |
5,980 |
Bangladesh |
1999-2000 |
44,726 |
9,854 |
Benin |
1996 (2001) |
2,115 |
5,796 |
Bolivia |
1998 |
2,601 |
12,109 |
Brazil |
1996 |
50,868 |
13,283 |
Burkina Faso |
1998-1999 |
3,976 |
4,812 |
Cameroon |
1998 |
4,996 |
4,697 |
Central African Republic |
1994-1995 |
1,199 |
5,551 |
Chad |
1996-1997 |
2,491 |
6,840 |
Colombia |
2000 |
12,346 |
10,907 |
Comoros |
1996 |
240 |
2,252 |
Côte d’Ivoire |
1998-1999 |
5,595 |
2,122 |
Dominican Republic |
1996 (1999) |
2,603 |
8,831 |
Egypt |
2000 |
21,991 |
16,957 |
Ethiopia |
1999 |
19,988 |
14,072 |
Gabon |
2000 |
348 |
6,203 |
Ghana |
1998-1999 |
6,581 |
6,003 |
Guatemala |
1998-1999 |
3,830 |
5,587 |
Guinea |
1999 |
2,637 |
5,090 |
Haiti |
2000 |
2,881 |
9,595 |
India |
1998-2000 |
298,291 |
92,486 |
Indonesia |
1997 |
64,059 |
34,255 |
Jordan |
1997 |
1,610 |
7,335 |
Kazakhstan |
1999 |
4,631 |
5,844 |
Kenya |
1998 |
11,306 |
8,380 |
Kyrgyz Republic |
1997 |
1,533 |
3,672 |
Female Sample Size 15-49 |
Male Sample Size* |
Chapters in Which Data Were Used |
|||
Education |
Health |
Marriage |
Parenthood |
||
6,430 |
1,719 |
Y |
Y |
Y |
Y |
10,544** |
2,556 |
Y |
|
Y |
Y |
6,219 |
2,709 |
Y |
Y |
Y |
Y (2001) |
11,187 |
3,780 |
Y |
Y |
Y |
Y |
12,612 |
2,949 |
Y |
Y |
Y |
Y |
6,445 |
2,641 |
Y |
Y |
Y |
Y |
5,501 |
2,562 |
Y |
Y |
Y |
Y |
5,884 |
1,729 |
Y |
Y |
Y |
Y |
7,454 |
2,320 |
Y |
Y |
Y |
Y |
11,585 |
|
Y |
Y |
Y |
Y |
3,050 |
795 |
Y |
Y |
Y |
Y |
3,040 |
886 |
Y |
Y |
Y |
Y |
8,422 |
2,279 |
Y |
Y |
Y |
Y |
15,573 |
|
Y |
|
Y |
Y |
15,367 |
2,607 |
Y |
Y |
Y |
Y |
6,183 |
2,004 |
|
|
Y |
|
4,843 |
1,546 |
Y |
Y |
Y |
Y |
6,021 |
|
Y |
Y |
Y |
Y |
6,753 |
1,980 |
Y |
Y |
Y |
Y |
10,159 |
3,171 |
Y |
Y |
Y |
Y |
90,303 |
|
Y |
|
Y |
Y |
28,810 |
Y |
Y |
Y |
||
5,548 |
Y |
Y |
Y |
||
4,800 |
1,440 |
Y |
Y |
Y |
Y |
7,881 |
3,407 |
Y |
Y |
Y |
Y |
3,848 |
|
Y |
Y |
Y |
Y |
Female Sample Size 15-49 |
Male Sample Size* |
Chapters in Which Data Were Used |
|||
Education |
Health |
Marriage |
Parenthood |
||
7,060 |
|
Y |
Y |
Y |
Y |
13,220 |
3,092 |
Y |
Y |
Y |
Y |
12,817 |
3,390 |
Y |
Y |
Y |
Y |
9,256 |
1,336 |
Y |
|
Y |
Y |
8,779 |
2,335 |
Y |
Y |
Y |
Y |
5,421 |
|
Y |
Y |
Y |
Y |
8,726 |
2,261 |
Y |
|
Y |
Y |
13,634 |
2,912 |
Y |
Y |
Y |
Y (2001) |
7,577 |
3,542 |
Y |
Y |
Y |
Y |
9,810** |
2,680 |
Y |
Y |
Y |
Y |
6,611 |
1,354 |
Y |
|
Y |
Y |
5,827 |
|
|
Y |
Y |
Y |
27,843 |
|
Y |
Y |
Y |
Y |
13,983 |
|
Y |
|
Y |
Y |
10,421 |
2,717 |
Y |
Y |
Y |
Y |
6,310 |
1,436 |
Y |
Y (1997) |
Y (1997) |
Y (1997) |
11,735 |
|
Y |
Y |
Y |
Y |
8,569 |
3,819 |
Y |
Y |
Y |
Y |
8,576 |
1,971 |
|
Y |
Y |
Y |
7,246 |
1,962 |
Y |
Y |
Y |
Y |
4,029 |
6,000 |
Y |
|
Y |
Y |
4,415 |
|
Y |
Y |
Y |
Y |
5,664 |
|
Y |
|
Y |
Y |
5,687 |
|
|
|
Y |
Y |
8,021 |
1,849 |
Y |
Y |
Y |
Y |
5,907 |
2,609 |
Y |
Y |
Y |
Y |
TABLE A-2 Demographic and Health Surveys (DHS)
Country |
Sorted by Region |
||
Regiona |
Most Recent Survey |
* = includes Male Survey |
|
Dominican Republic |
Carib/CA |
1996 |
* |
Guatemala |
Carib/CA |
1998-1999 |
|
Haiti |
Carib/CA |
2000 |
* |
Nicaragua |
Carib/CA |
1997-1998 |
* |
Comoros |
E/S Africa |
1996 |
* |
Ethiopia |
E/S Africa |
1999 |
* |
Kenya |
E/S Africa |
1998 |
* |
Madagascar |
E/S Africa |
1997 |
|
Malawi |
E/S Africa |
2000 |
* |
Mozambique |
E/S Africa |
1997 |
* |
Namibia |
E/S Africa |
1992 |
|
Rwanda |
E/S Africa |
2000 |
|
South Africa |
E/S Africa |
1998 |
|
Tanzania |
E/S Africa |
1999 |
* |
Uganda |
E/S Africa |
2000-2001 |
* |
Zambia |
E/S Africa |
1996-1997 |
* |
Zimbabwe |
E/S Africa |
1999 |
* |
Egypt |
ME |
2000 |
|
Jordan |
ME |
1997 |
|
Morocco |
ME |
1992 |
|
Turkey |
ME |
1998 |
*b |
Yemen |
ME |
1991-1992 |
|
Bolivia |
SA |
1998 |
* |
Brazil |
SA |
1996 |
* |
Colombia |
SA |
2000 |
|
Paraguay |
SA |
1990 |
|
Peru |
SA |
2000 |
* |
Bangladesh |
SC/SE Asia |
1999-2000 |
|
India |
SC/SE Asia |
1998-2000 |
|
Indonesia |
SC/SE Asia |
1997 |
|
Nepal |
SC/SE Asia |
2000-2001 |
Country |
Sorted by World Bank Income Categories |
||
Income |
Most Recent Survey |
* = includes Male Survey |
|
Armenia |
Low |
2000 |
*b |
Bangladesh |
Low |
1999-2000 |
|
Benin |
Low |
1996 |
* |
Burkina Faso |
Low |
1998-1999 |
* |
Cameroon |
Low |
1998 |
* |
Central African Republic |
Low |
1994-1995 |
* |
Chad |
Low |
1996-1997 |
* |
Comoros |
Low |
1996 |
* |
Côte d’Ivoire |
Low |
1998-1999 |
* |
Ethiopia |
Low |
1999 |
* |
Ghana |
Low |
1998-1999 |
* |
Guinea |
Low |
1999 |
* |
Haiti |
Low |
2000 |
* |
India |
Low |
1998-2000 |
|
Indonesia |
Low |
1997 |
|
Kenya |
Low |
1998 |
* |
Kyrgyz Republic |
Low |
1997 |
|
Madagascar |
Low |
1997 |
|
Malawi |
Low |
2000 |
* |
Mali |
Low |
2001 |
* |
Mozambique |
Low |
1997 |
* |
Nepal |
Low |
2000-2001 |
|
Nicaragua |
Low |
1997-1998 |
* |
Niger |
Low |
1998 |
* |
Nigeria |
Low |
1999 |
* |
Pakistan |
Low |
1990-1991 |
|
Rwanda |
Low |
2000 |
|
Senegal |
Low |
1997 |
* |
Tanzania |
Low |
1999 |
* |
Togo |
Low |
1998 |
* |
Uganda |
Low |
2000-2001 |
* |
Country |
Sorted by Region |
||
Regiona |
Most Recent Survey |
* = includes Male Survey |
|
Pakistan |
SC/SE Asia |
1990-1991 |
|
Philippines |
SC/SE Asia |
1998 |
|
Vietnam |
SC/SE Asia |
1997 |
|
Armenia |
Soviet |
2000 |
*b |
Kazakhstan |
Soviet |
1999 |
*b |
Kyrgyz Republic |
Soviet |
1997 |
|
Uzbekistan |
Soviet |
1996 |
|
Benin |
W/M Africa |
1996 |
* |
Burkina Faso |
W/M Africa |
1998-1999 |
* |
Cameroon |
W/M Africa |
1998 |
* |
Central African Republic |
W/M Africa |
1994-1995 |
* |
Chad |
W/M Africa |
1996-1997 |
* |
Cote d’Ivoire |
W/M Africa |
1998-1999 |
* |
Gabon |
W/M Africa |
2000 |
*c |
Ghana |
W/M Africa |
1998-1999 |
* |
Guinea |
W/M |
Africa 1999 |
* |
Mali |
W/M Africa |
2001 |
* |
Niger |
W/M Africa |
1998 |
* |
Nigeria |
W/M Africa |
1999 |
* |
Senegal |
W/M Africa |
1997 |
* |
Togo |
W/M Africa |
1998 |
* |
aKey: Carib/CA (Caribbean and Central America); E/S Africa (Eastern and Southern Africa); ME (Middle East [Northern Africa and Western Asia]); SA (South America); SC/SE Asia (South-central and South-eastern Asia); Soviet (Former Soviet Asia); W/M Africa (Western and Middle Africa). bMale survey data are available for these countries, but not in sufficient number to allow aggregation of data to generate regional averages. |
for making cross-national comparisons. Annual enrollment data provided by UNESCO are based on enrollments as reported officially by schools to national ministries of education. These annual enrollment counts are divided by United Nations estimates for the population for the year and ages in question to derive gross or net enrollment ratios for each level of schooling. Gross enrollment ratios, which are available for almost all countries for
Country |
Sorted by World Bank Income Categories |
||
Income |
Most Recent Survey |
* = includes Male Survey |
|
Uzbekistan |
Low |
1996 |
|
Vietnam |
Low |
1997 |
|
Yemen |
Low |
1991-1992 |
|
Zambia |
Low |
1996-1997 |
* |
Zimbabwe |
Low |
1999 |
* |
Bolivia |
Lower middle |
1998 |
* |
Colombia |
Lower middle |
2000 |
|
Dominican Republic |
Lower middle |
1996 |
* |
Egypt |
Lower middle |
2000 |
|
Guatemala |
Lower middle |
1998-1999 |
|
Jordan |
Lower middle |
1997 |
|
Kazakhstan |
Lower middle |
1999 |
*b |
Morocco |
Lower middle |
1992 |
|
Namibia |
Lower middle |
1992 |
|
Paraguay |
Lower middle |
1990 |
|
Peru |
Lower middle |
2000 |
* |
Philippines |
Lower middle |
1998 |
|
Brazil |
Upper middle |
1996 |
* |
Gabon |
Upper middle |
2000 |
*c |
South Africa |
Upper middle |
1998 |
|
Turkey |
Upper middle |
1998 |
*b |
cGabon data on women unavailable at time of this analysis; data on men do not include schooling. NOTE: Middle East, South-central, and South-eastern Asia are excluded from marriage chapter Table 7-9 because the surveys are based on ever-married samples. |
multiple decades, relate enrollment in primary or secondary levels regardless of age to the population age group appropriate to each level of schooling. Net enrollment ratios, which are confined to those enrolled in the eligible age range, are conceptually cleaner but less widely available particularly for earlier years, because many countries have not collected data on enrollment by age. Because these data are school based, they have
the potential to be broken down into geographic subgroupings but not by household characteristics, such as wealth.
These enrollment data vary in quality according to the quality of the management information systems in each country. The development of good systems is a continuing challenge in many parts of the developing world (Moulton et al., 2001). School reform efforts, which often include improvements in management information systems that make current data more accurate, may compromise comparability over time. Furthermore, in some settings in which financial flows to schools are a function of the level of enrollment, there can be a substantial motivation to inflate reported enrollments.
Some of the shortcomings of UNESCO enrollment data have been discussed in the literature (Behrman and Rosenzweig, 1994; Hewett and Lloyd, 2005; Lloyd, Mensch, and Clark, 2000). While some researchers feel comfortable relying on the enrollment data from UNESCO for trends and cross-country comparisons (e.g., Behrman and Sengupta, 2002), others doubt that the inherent biases in the data are sufficiently consistent across countries or over time to permit firm comparative conclusions about levels, trends, and differentials (Hewett and Lloyd, 2005; Lloyd et al., 2000).
With the establishment of the UNESCO Institute of Statistics (UIS) in 2001, the international community has taken some important steps to improve systems of reporting on literacy and schooling, including launching several special initiatives with the Organisation for Economic Co-operation and Development (OECD) and the World Bank to strengthen the collection and reporting of comparative statistics and indicators. The World Education Indicators program (WEI) launched in 1997 is one example. This program now includes 17 developing countries (Argentina, Brazil, China, Egypt, India, Indonesia, Jamaica, Jordan, Malaysia, Paraguay, Peru, the Philippines, the Russian Federation, Thailand, Tunisia, Uruguay, and Zimbabwe). Eventually this program will permit comparisons between developed and developing country data on a more consistent basis. Recently, Bruns, Mingat, and Rakotomalala (2003) from the World Bank have used UNESCO data to calculate a primary school completion ratio using data from UNESCO on end of year enrollments in the last year of primary school (if available) or beginning-of-year enrollment data for the last year of primary school adjusted for repeaters. These estimates of numbers graduating from primary school are divided by UN estimates of the population of official graduation age to simulate a completion rate and are available now to assess changes in the 1990s.
Relative to data on enrollment from UNESCO, data on school attendance and attainment derived from censuses and national sample surveys have some advantages and some disadvantages. The major advantages in-
clude (1) the fact that the data for the numerator and the denominator of any indicator are based on the same underlying population, thus allowing the construction of proper cohort-specific indicators on a consistent basis over time, (2) that data collected from households are more likely to capture current attendance rather than opening-day enrollment, thus providing a much more realistic measure of school participation, and (3) that attendance rates from household data can be compared by household income or wealth groups as well as other household characteristics. The disadvantages of household data include the facts that (1) they often are only collected periodically rather than annually, (2) differential mortality across education groups can bias the estimation of trends,4 and (3) in the case of household surveys, changing sample frames over time can compromise the comparability of successive surveys. Despite their advantages, household survey data rarely have been used for comparative analysis of educational trends, because of their lack of accessibility on a comparable basis. There has never been an international education survey program similar to the DHS program. Census data are published with a huge lag and, despite much UN technical assistance over the years, often lack comparability when presented in tabular form in printed reports (Lloyd et al., 2000).5 The fact that DHS data were collected on schooling on a comparable basis has allowed us to take advantages of some of the benefits offered by household data for the study of trends in schooling across countries and regions. See Table A-1 for the list of countries included in the presentation of data on schooling.
The data presented on standardized text scores come primarily from Programme for International Student Assessment (PISA) of the Organisation for Economic Co-operation and Development (OECD) (Organisation for Economic Co-operation and Development, 2001) and from the Third International Mathematics and Science Study (TIMSS) undertaken in 1999 by the International Study Center at Boston College—both international efforts with some participation of developing countries.
4 |
When using cross-sectional cohort data to estimate trends, there is the danger that growth in schooling attainment will be underestimated if there is differential mortality by schooling in the relevant age groups. Given that we are comparing age groups between 10-14 and 30-34, for which group mortality rates are relatively low, this is unlikely to be a serious source of bias (Lloyd et al., 2000). |
5 |
There are several current efforts to make census data more accessible: Integrated Public Use Microdata Series (IPUMS, http://www.ipums.umn.edu) based at the University of Michigan and the African Census Analysis Project based at the University of Pennsylvania (http://www.acap.upenn.edu). |
Chapter 4: Health
The panel relied on multiple sources of data for the health chapter, depending on the topic. Data on mortality and morbidity were taken from WHO and the Joint United Nations Programme on HIV/AIDS. DHS surveys were used to analyze data on sexual initiation and on contraceptive use (see Table A-1). While data on sex and contraceptive use among young people have also been collected by the Centers for Disease Control and Prevention as part of their Young Adult Reproductive Health Survey, relatively few developing countries (as defined by the panel) have participated in these surveys and the data are not as easily accessible.6 The global tobacco surveys were used to describe levels and trends in smoking worldwide. For some topics, regional and global averages were not constructed, either because coverage was limited or because data were available for an insufficient number of countries.
Chapter 5: Work
The panel relied on data from labor force surveys provided by the International Labour Organization’s LABORSTAT (http://laborsta.ilo.org/ from November 24, 2003) trends in labor force participation rates, data on population trends from the UN Population Division, and data collected by the Population Council on time use. To analyze trends in transitions to education and to work for seven case study countries, the panel used large public-use samples from the two most recent censuses for Mexico (1990 and 2000), Kenya (1989 and 1999), and Vietnam (1989 and 1999), which are available from the IPUMS-International web site at the University of Minnesota (www.ipums.umn.edu), as well as the 1992 and 1999 surveys of Brazil’s Pesquisa Nacional por Amostra de Domicilios (PNAD); the 1993 South Africa Integrated Household Survey, SALDRU/World Bank and September 2000 South Africa Labour Force Survey, Statistics South Africa, courtesy of David Lam (University of Michigan); the Child Health and Nutrition Survey in China from 1989 and 1997 (http://www.cpc.edu/dataarch/primary), courtesy of Emily Hannum (University of Pennsylvania); and the 1987 Social and Economic Survey of Households, Statistical Center of Iran, and the 1998 Household Expenditure and Income Survey, Statistical Center of Iran, courtesy of Djavad Salehi-Isfahani (Virginia Polytechnic University). These data were also used in Chapter 7.
Other data, typically using countries as examples, were drawn from previously published analyses and used to illustrate various points.
Chapter 6: Citizenship
After an initial exploration of a range of data sets including the World Values Surveys, AfroBarometer surveys, LatinoBarometer surveys, and UNICEF young voices surveys, the panel decided to rely primarily on data from the World Values Surveys collected for the first time in a large sample of developing countries in 1995-1998. This survey project is guided by a steering committee representing all regions of the world. Coordination and distribution of the data are based at the Institute for Social Research at the University of Michigan under the direction of Ronald Inglehart. While this choice meant that we have very little coverage for Africa (except Nigeria, which has about a fifth of the continent’s population), we thought that the sampling procedures for the World Values Surveys were better documented than were alternative sources, such as the Afro- and Latino-Barometer surveys. Because of small sample sizes, however, we thought we needed to allow larger age ranges (18-34, 35+) in order to have a sufficient sample size to explore gender differences. These data were supplemented for some topics by published data from the UNICEF young voices surveys.
More recent versions of the World Values Surveys (1999-2000) and AfroBarometer Surveys (1999-2000) are better documented and include more developing countries; however, neither of these newer data sets were publicly available at the time that this chapter was being completed.
Chapter 7: Marriage
In order to analyze marriage patterns, in particular trends in age at first marriage, it would be desirable to have accurate marriage registration data for at least two points in time. Such data are rarely available for developing countries. The United Nations Population Division collects data on the percentage of the population married in five-year age groups for most developing countries. For the most part, these data are available for men as well as women. For this analysis, we consider all countries in Africa, Asia, and Latin America and the Caribbean, with the exception of those identified by the World Bank as “high income” and those with a population under 140,0007 in population (World Bank, 2002b). Given the focus on trends, we have identified 73 countries of the 117 that meet our criteria for which recent data,8 i.e., data collected in 1990 or later, are available and for which there is information from two censuses or surveys at least 10 years apart (see Table A-3). In 2000, the United Nations estimated that
TABLE A-3 United Nations Database on Marriage
Sorted by Region |
Sorted by Region Categories |
||
Country |
Regiona |
Census/Survey Year 1 |
Census/Survey Year 2 |
Belize |
Carib/CA |
1980 |
1991 |
Dominican Republic |
Carib/CA |
1981 |
1996 |
El Salvador |
Carib/CA |
1971 |
1992 |
Guatemala |
Carib/CA |
1973 |
1990 |
Haiti |
Carib/CA |
1989 |
2000 |
Mexico |
Carib/CA |
1980 |
1990 |
Nicaragua |
Carib/CA |
1971 |
1998 |
Panama |
Carib/CA |
1980 |
1990 |
Puerto Rico |
Carib/CA |
1980 |
1990 |
Trinidad and Tobago |
Carib/CA |
1980 |
1990 |
Botswana |
E/S Africa |
1981 |
1991 |
Burundi |
E/S Africa |
1979 |
1990 |
Comoros |
E/S Africa |
1980 |
1996 |
Ethiopia |
E/S Africa |
1984 |
2000 |
Kenya |
E/S Africa |
1969 |
1998 |
Malawi |
E/S Africa |
1987 |
2000 |
Mauritius |
E/S Africa |
1972 |
1990 |
Mozambique |
E/S Africa |
1980 |
1997 |
Namibia |
E/S Africa |
1960 |
1991 |
Rwanda |
E/S Africa |
1978 |
1996 |
South Africa |
E/S Africa |
1985 |
1996 |
Tanzania |
E/S Africa |
1978 |
1996 |
Uganda |
E/S Africa |
1969 |
1995 |
Zambia |
E/S Africa |
1980 |
1999 |
Zimbabwe |
E/S Africa |
1982 |
1999 |
China |
EA |
1987 |
1999 |
Bahrain |
ME |
1981 |
1991 |
Egypt |
ME |
1986 |
1996 |
Jordan |
ME |
1979 |
1994 |
Morocco |
ME |
1982 |
1994 |
Occ. Palestinian Territory |
ME |
1967 |
1997 |
Sudan |
ME |
1983 |
1993 |
Tunisia |
ME |
1984 |
1994 |
Turkey |
ME |
1980 |
1990 |
Argentina |
SA |
1980 |
1991 |
Bolivia |
SA |
1988 |
1998 |
Brazil |
SA |
1980 |
1996 |
Chile |
SA |
1982 |
1992 |
Colombia |
SA |
1973 |
1993 |
Ecuador |
SA |
1974 |
1990 |
Guyana |
SA |
1980 |
1991 |
Paraguay |
SA |
1982 |
1992 |
Peru |
SA |
1981 |
1996 |
Sorted by World Bank Income Categories |
|||
U.N. Marriage Data Country |
Income |
Census/Survey Year 1 |
Census/Survey Year 2 |
Azerbaijan |
Low |
1989 |
1999 |
Bangladesh |
Low |
1981 |
1991 |
Benin |
Low |
1979 |
1996 |
Burkina Faso |
Low |
1985 |
1999 |
Burundi |
Low |
1979 |
1990 |
Cambodia |
Low |
1962 |
1998 |
Cameroon |
Low |
1987 |
1998 |
Central African Republic |
Low |
1975 |
1994-1995 |
Chad |
Low |
1964 |
1996 |
Comoros |
Low |
1980 |
1996 |
Côte d’Ivoire |
Low |
1978 |
1994 |
Ethiopia |
Low |
1984 |
2000 |
Gambia |
Low |
1983 |
1993 |
Haiti |
Low |
1989 |
2000 |
India |
Low |
1981 |
1992-1993 |
Indonesia |
Low |
1980 |
1990 |
Kenya |
Low |
1969 |
1998 |
Kyrgyz Republic |
Low |
1989 |
1999 |
Malawi |
Low |
1987 |
2000 |
Mali |
Low |
1976 |
1995-1996 |
Mauritania |
Low |
1988 |
2000-2001 |
Mozambique |
Low |
1980 |
1997 |
Myanmar |
Low |
1973 |
1991 |
Nepal |
Low |
1981 |
1991 |
Nicaragua |
Low |
1971 |
1998 |
Niger |
Low |
1988 |
1998 |
Pakistan |
Low |
1981 |
1998 |
Rwanda |
Low |
1978 |
1996 |
Senegal |
Low |
1978 |
1997 |
Sudan |
Low |
1983 |
1993 |
Tanzania |
Low |
1978 |
1996 |
Uganda |
Low |
1969 |
1995 |
Zambia |
Low |
1980 |
1999 |
Zimbabwe |
Low |
1982 |
1999 |
Belize |
Lower Middle |
1980 |
1991 |
Bolivia |
Lower Middle |
1988 |
1998 |
Cape Verde |
Lower Middle |
1980 |
1990 |
China |
Lower Middle |
1987 |
1999 |
Colombia |
Lower Middle |
1973 |
1993 |
Dominican Republic |
Lower Middle |
1981 |
1996 |
Ecuador |
Lower Middle |
1974 |
1990 |
Egypt |
Lower Middle |
1986 |
1996 |
El Salvador |
Lower Middle |
1971 |
1992 |
Sorted by Region |
Sorted by Region Categories |
||
Country |
Regiona |
Census/Survey Year 1 |
Census/Survey Year 2 |
Uruguay |
SA |
1985 |
1996 |
Venezuela |
SA |
1974 |
1990 |
Bangladesh |
SC/SE Asia |
1981 |
1991 |
Cambodia |
SC/SE Asia |
1962 |
1998 |
India |
SC/SE Asia |
1981 |
1992-1993 |
Indonesia |
SC/SE Asia |
1980 |
1990 |
Iran |
SC/SE Asia |
1986 |
1996 |
Malaysia |
SC/SE Asia |
1980 |
1991 |
Maldives |
SC/SE Asia |
1985 |
1995 |
Myanmar |
SC/SE Asia |
1973 |
1991 |
Nepal |
SC/SE Asia |
1981 |
1991 |
Pakistan |
SC/SE Asia |
1981 |
1998 |
Philippines |
SC/SE Asia |
1980 |
1995 |
Thailand |
SC/SE Asia |
1980 |
1990 |
Azerbaijan |
Soviet |
1989 |
1999 |
Kazakhstan |
Soviet |
1989 |
1999 |
Kyrgyz Republic |
Soviet |
1989 |
1999 |
Benin |
W/M Africa |
1979 |
1996 |
Burkina Faso |
W/M Africa |
1985 |
1999 |
Cameroon |
W/M Africa |
1987 |
1998 |
Cape Verde |
W/M Africa |
1980 |
1990 |
Central African Republic |
W/M Africa |
1975 |
1994-1995 |
Chad |
W/M Africa |
1964 |
1996 |
Côte d’Ivoire |
W/M Africa |
1978 |
1994 |
Gabon |
W/M Africa |
1961 |
2000 |
Gambia |
W/M Africa |
1983 |
1993 |
Mali |
W/M Africa |
1976 |
1995-1996 |
Mauritania |
W/M Africa |
1988 |
2000-2001 |
Niger |
W/M Africa |
1988 |
1998 |
Senegal |
W/M Africa |
1978 |
1997 |
aKey: Carib/CA (Caribbean and Central America); E/S Africa (Eastern and Southern Africa); ME (Middle East [Northern Africa and Western Asia]); SA (South America); SC/SE Asia (South-central and South-eastern Asia); Soviet (Former Soviet Asia); W/M Africa (Western and Middle Africa). |
there were 1.4 billion young people ages 10-24 in these 117 countries; 87 percent or 1.2 billion were resident in the 73 countries for which marriage trend data are available. Because of the difficulty of interpreting such large volumes of information, we have aggregated these data by UN geographic groupings.
Sorted by World Bank Income Categories |
|||
U.N. Marriage Data Country |
Income |
Census/Survey Year 1 |
Census/Survey Year 2 |
Guatemala |
Lower Middle |
1973 |
1990 |
Guyana |
Lower Middle |
1980 |
1991 |
Iran |
Lower Middle |
1986 |
1996 |
Jordan |
Lower Middle |
1979 |
1994 |
Kazakhstan |
Lower Middle |
1989 |
1999 |
Maldives |
Lower Middle |
1985 |
1995 |
Morocco |
Lower Middle |
1982 |
1994 |
Namibia |
Lower Middle |
1960 |
1991 |
Occ. Palestinian Territory |
Lower Middle |
1967 |
1997 |
Paraguay |
Lower Middle |
1982 |
1992 |
Peru |
Lower Middle |
1981 |
1996 |
Philippines |
Lower Middle |
1980 |
1995 |
Thailand |
Lower Middle |
1980 |
1990 |
Tunisia |
Lower Middle |
1984 |
1994 |
Argentina |
Upper Middle |
1980 |
1991 |
Bahrain |
Upper Middle |
1981 |
1991 |
Botswana |
Upper Middle |
1981 |
1991 |
Brazil |
Upper Middle |
1980 |
1996 |
Chile |
Upper Middle |
1982 |
1992 |
Gabon |
Upper Middle |
1961 |
2000 |
Malaysia |
Upper Middle |
1980 |
1991 |
Mauritius |
Upper Middle |
1972 |
1990 |
Mexico |
Upper Middle |
1980 |
1990 |
Panama |
Upper Middle |
1980 |
1990 |
Puerto Rico |
Upper Middle |
1980 |
1990 |
South Africa |
Upper Middle |
1985 |
1996 |
Trinidad and Tobago |
Upper Middle |
1980 |
1990 |
Turkey |
Upper Middle |
1980 |
1990 |
Uruguay |
Upper Middle |
1985 |
1996 |
Venezuela |
Upper Middle |
1974 |
1990 |
Coverage varies considerably by region, with approximately 90 percent or more of the population represented by these data in Eastern and Southern Africa, South-central and South-eastern Asia, Eastern Asia, South America, and the Caribbean and Central America, but only 63 percent represented in the Middle East, 31 percent represented in Western and
Middle Africa, and 38 percent represented in the former Soviet Asia. Note that Eastern Asia consists entirely of China, as data are unavailable for the two other countries, Mongolia and North Korea. Populous countries for which data are unavailable include Afghanistan, the Democratic Republic of the Congo, Iraq, Nigeria, Saudi Arabia, Uzbekistan, and Vietnam.
The Demographic and Health Surveys provide additional information to what is available from the United Nations data base (see Tables A-1 and A-2). Respondents’ specific age at first marriage is obtained on these surveys, enabling the calculation of the proportion married by a particular age rather than just the percentage of a particular group who are married. In addition, they enable one to examine differentials in the timing of marriage by schooling attainment, place of residence and household economic status and may therefore provide insights into the forces behind the changes we have observed. The one drawback is that the surveys have been conducted in fewer countries than we have UN data for.
Note that regions vary considerably in the number of countries for which DHS surveys have been conducted. Coverage is highest in Eastern and Southern Africa, with approximately 92 percent of the population represented, and lowest in the Caribbean and Central America, where, because no recent survey is available for Mexico—by far the largest country—only about one-fifth of the population is represented. It is also important to keep in mind that no data are available for Eastern Asia, which includes China. Note, however, there are a few countries for which DHS data are available that are not included in the UN database: Armenia, Ghana, Guinea, Madagascar, Nigeria, Togo, Uzbekistan, Vietnam, and Yemen). Indeed, for two regions, Western/Middle Africa and Former Soviet Asia, coverage is considerably higher in the DHS at 75 percent and 68 percent, respectively. As with the UN data, the regional analyses based on the DHS data are weighted averages.
Chapter 8: Parenthood
All data, for the parenthood chapter, are based on DHS surveys (see earlier discussion of DHS data as well as Tables A-1 and A-2).