Advocacy in epidemiology and demography.

This paper is a summary of a panel discussion at the Conference on Epidemiology and Demography held at Georgetown University, in Washington D.C. on February 8-9, 2001. The participants were Al Hermalin, Linda Martin, Mike Stoto, Robert Wallace, Douglas Weed, and Rose Li (who chaired the session). A list of questions similar to the section headings in this paper was prepared in advance of the conference, and each of the participants was asked to address specific issues, although the presentations typically covered a range of topics. This summary also includes comments from the floor.

Greater collaboration across the disciplines: challenges and opportunities.

This paper reports a panel discussion--Opportunities for and Limitations to Greater Collaboration Across the Disciplines--held at the conference. It highlights the need for greater collaboration between demographers and epidemiologists and notes the institutional and disciplinary challenges to and opportunities for promoting greater cooperation.

The effect of education on mortality among older Taiwanese and its pathways.

The present research examines the impact of education on the mortality of older Taiwanese during a 4-year interval from April 1989 to April 1993. Data used for this study come from the Taiwan Survey of Health & Living Status of the Elderly (1989). The research decomposes the effect of education into the direct effect and the indirect effects by means of health status, health behaviors, and social relationships. We have shown that, of the total effect of educational attainment on the mortality of older Taiwanese, about 83% represents indirect influences by means of the 3 mediating factors, particularly health status. On the other hand, the magnitude of the direct effect, which might reflect influences of additional intervening variables on old-age mortality, is low and not statistically significant. The results demonstrate that the apparent strong effect of education on mortality among older Taiwanese can be accounted for parsimoniously through 3 major pathways.

Drawing policy lessons for Asia from research on ageing.

PIP: Many countries in the ESCAP region have led the way among developing and newly industrializing countries in anticipating the consequences of population aging, and are looking for feasible, affordable ways to cope. In East and Southeast Asia, there have been many conferences, research projects, books, and papers on population aging since the early 1980s, even though only Hong Kong and Japan had more than 10% of their populations aged 60 years or older as of 1990. At least one population-related survey has been conducted in every country in the region; the major emphases have been upon living arrangements, informal support patterns, and certain dimensions of health. The ESCAP region's early attention to population aging has given it a strong foundation upon which to base future efforts. Developing a conceptual framework for policy and research, and implications for policy and program development are discussed.^ieng

A demographic decomposition of elderly living arrangements with a Mexican example.

Population trends in developing countries raise concern about support for the elderly. The proportion of elderly living with extended kin is an indicator of support. This paper considers the analytic utility of a demographic decomposition of living arrangements of elderly Mexicans into population components which include weights for age and marital composition and corresponding rates or propensities. Separate decompositions for elderly males and females demonstrate the importance of population composition to the makeup of the elderly population who are living with extended kin. The utility of the decomposition for comparative analysis is demonstrated by decomposing gender differences in living arrangements. The higher proportion of women living with extended kin is primarily the result of gender differences in age-specific marital status and only secondarily the results of actual differences in propensities toward this type of living arrangement. The utility and limitations of this analytic tool for comparative research are discussed.

Census-based approaches for studying aggregate changes in characteristics of the elderly.

"This article illustrates a few simple techniques and approaches for using census data on the elderly more completely." The authors find that "by fully exploiting census and similar data, it is possible for analysts to gain insights about the future composition of the elderly, the changing status of the elderly in relation to younger groups, and the rate at which key cohorts are changing in various characteristics." The geographical focus is on developing countries.

Do statements about desired family size predict fertility? The case of Taiwan, 1967-1970.

The predictive accuracy of respondents' statements about their future fertility is examined, using interview data from a longitudinal study conducted in Taiwan. Two measures of preference are found to be highly intercorrelated; and regardless of which one is used, Taiwanese women are shown to predict their subsequent fertility at least as well as U.S. women. The preference measures are also predictive of rates of contraceptive use and abortion. While demographic and social characteristics are correlated with fertility in expected directions, statements about wanting more children prove to be highly predictive of subsequent fertility for both modern and less advanced segments of the population.

Effects of birth rank, maternal age, birth interval, and sibship size on infant and child mortality: evidence from 18th and 19th century reproductive histories.

There has been long-standing interest in the effects of maternal age, birth rank, and birth spacing on infant and child mortality. Contradictory inferences about the role of these factors have arisen on occasion because of the absence of adequate controls, the use of cross-sectional or incomplete reproductive histories, and inattention to the effect of family size goals and birth limitation practices. This study analyzes completed reproductive histories for German village populations in the 18th and 19th centuries, a period when deliberate fertility control was largely absent. Our results confirm previous studies of the association of infant mortality with maternal age, although in the present data these differentials are largely limited to neonatal mortality. They also confirm the importance of birth interval as a factor in infant mortality. Sibship size is positively related to infant mortality even when birth rank is controlled. However, once sibship size is controlled, there are no systematic differences in infant and child mortality by birth order. The mechanisms relating sibship size and mortality are explored.

Motivational factors in IUD termination: data from the second Taiwan IUD follow-up survey.

PIP: In order to determine whether or not they were still using the IUD or their reasons for discontinuation if discontinuation had occurred, a sample of 4648 women who had accepted an IUD between March 1965-June 1966 were interviewed between December 1966-February 1977. The results are very similar to those obtained from an earlier survey and from other surveys taken in Taiwan. They showed: 1) the median duration of use is about 20 months, 2) about 1/3 of all terminations are because of the device itself (pregnancy or expulsion) and 2/3 are because of removal, and 3) the rates of removal do not stabilize and level off at about 12 months but continue to rise. The latter is an unexpected result; the expectation had been that most of the physical side effects of the IUD would subside after 6-12 months of use. It needs to be emphasized that IUD termination is not simply a matter of medical and physiological factors related to the IUD itself. Motivational factors also seem to be extremely important. An assessment of the motivational factors was made by exploring gross cumulative termination rates according to reason for using family planning (limitation vs. spacing), reason for removal (medical vs. nonmedical), and comparison of number of living children with number of chidren desired. In every case motivational factors are associated with continuation rate: 1) women who accept family planning to terminate childbearing have lower rates of removal for nonmedical reasons than women who accept family planning for spacing, 2) women who have more children than they want have lower rates of termination than women who have the desired number, and 3) women who have the desired number have lower termination rates than women who have fewer children than they want.^ieng

Birth spacing and infant mortality: evidence for eighteenth and nineteenth century German villages.

Data from an historical population in which fertility control was minimal and modern health services were mostly unavailable are used to show that there appears to have been a strong association between previous birth interval length and infant mortality, especially when the previous child survived. Although only imperfect proxies for breast-feeding practices and other potentially confounding factors are available for this population, the results suggest that the association between previous interval length and infant mortality in this population is not solely, or primarily, a function of differences in breast-feeding behaviour or socioeconomic status. Other factors, e.g. maternal depletion or sibling competition, are more likely to explain the observed association.

PIP: In order to untangle the relationships between maternal age, birth spacing, family size, birth order and infant and child mortality, this study selected a data set which represented minimal fertility control and had no modern health services. 14 German villages with historical data from parish and civil registers on 48,000 births and 9000 reproductive histories were selected in this study of the relationship between reproductive patterns and infant mortality. The regions of Baden, Wurttemberg, Bavaria, Waldeck, and East Friesland represented different demographic conditions. The infant morality rate was 228/1000; there was wide variation in the prevalence and duration of breastfeeding. In East Friesland villages, breastfeeding was used by a majority of mothers for an average duration of a year, there is evidence to suggest that breastfeeding patterns are related to variations in infant mortality. Factors affecting the relationship between birth interval length and level of infant mortality and the nature of their effect are identified in the discussion and in chart form. 3 hypotheses are discussed: 1) short birth intervals do not permit adequate maternal recovery time; 2) the association between previous birth interval length and child survival is due to competition for food and care between the index child and the next oldest sibling; and 3) closely space children are more likely to spread infectious diseases to each other. The results showed that infant mortality risks declined as birth interval length increased before plateauing at 24-28 months. For children whose previous sibling died in infancy, there is not consistent relationship between birth interval length and infant mortality. In the multivariate analyses of those children whose next oldest sibling survived infancy, birth interval length is divided into 2 categories in order to capture the nonlinear relationship. Infant mortality rates were higher for older women, so age of the mother 35 was a variable. Analyses were conducted by birth order separately. In the logit analysis with maternal age and paternal occupation constant, for all birth orders with probability of dying decreases consistently as birth order increases. Older maternal age has no effect on birth orders higher than 6. Little association between occupation and mortality risks exists. Region and number of infant deaths are strongly related to infant mortality, where lower regions of breastfeeding had higher mortality. Regional variations are described in different analyses to confirm that ameliorative effects of breastfeeding on child survival, but differences in breastfeeding behavior are not solely responsible for interval length and mortality. Breastfeeding effects and previous infant deaths may reflect lower birth weight, poorer parental practices, and lower living standards. The results suggest that factors (breastfeeding and use of health services and family planning) may not be adequate to explain the relationship. Further understanding is required of the factors affecting interval length in order to make conclusions about the causal mechanisms.

Measuring accessibility to family planning services in rural Thailand.

There are a variety of ways to conceptualize and measure accessibility to contraceptive services and supplies. Using detailed data for rural Thailand, a multiplicity of reasonable measures are developed that reflect time and/or distance to various government program outlets. Many of these are only moderately correlated with one another, and to an important extent, different measures show different associations with levels of contraceptive prevalence. Clearly there is a general need in this area of investigation to pay more attention to the measurement of accessibility as a community characteristic and to consider the sensitivity of results to alternate measures.

PIP: A variety of ways to conceptualize and measure accessibility to contraceptive services and supplies are used to evaluate program accessibility in rural Thailand. Included are distance to a specific outlet, index of distances to various outlets, travel time to a specific outlet, index of travel times to various outlets, types of services provided, length of time that specified services and specified outlets have been available, convenience, in terms of ease and costs of transportation, quality of service in terms of waiting time and competence and attitude of the staff, and costs of family planning services. An alternative approach to measuring accessibility from data on individual respondents has been to develop community-level indicators. Thailand, the only country with detailed data collection on accessibility, has experienced rapid and pervasive changes in reproductive behavior during the last 2 decades. Contraceptive prevalence among married women of reproductive age increased from less than 15% in 1969 to almost 60% by 1981. This was more or less coincidental with the development of an active national program to promote family planning. A crucial aspect of the Thai program was the early implementation of a policy allowing nurses and auxiliary midwives to distribute oral contraceptives. Although costs vary by type of method and changes in prices, the cost structure is identical from village to village and is not, therefore, a matter of differential accessibility. A major conclusion is that there is a need for more attention to the measurement of accessibility and for further tests of the sensitivity of results to alternate measures. Particular focus should be given to capturing changes in accessibility over appropriate time periods, as well as measuring current configurations. In the future, it will become increasingly important to collect data on other dimensions of accessibility, e.g. the quality of services, the complement between the public and private sector and the extent of information, education, and communication (IEC) efforts.

A multilevel model of family planning availability and contraceptive use in rural Thailand.

This paper assesses the ways in which the availability of family planning program outlets influences the likelihood of contraceptive use in rural Thailand. It focuses on a village-level measure of actual availability of sources rather than respondent perceptions of availability. Individual-level and village-level data collected as part of the second Thailand Contraceptive Prevalence Survey are used to test three hypotheses about the effects of actual availability: that (a) availability of family planning outlets increases the likelihood of contraceptive use; (b) it enhances the effect of a desire for no more children on the likelihood of use; and (c) it weakens the positive relationship between education and the likelihood of use.

The impact of family planning clinic programs on adolescent pregnancy.

During the 1970s, there was a decline in adolescent childbearing in the United States and, among teenagers who were sexually active, there was a decline in pregnancy rates as well. To what extent was increased enrollment by teenagers in federally funded family planning clinics responsible for these declines? Areal multivariate analysis reveals that adolescent birthrates were reduced between 1970 and 1975 as the result of enrollment by teenagers in family planning clinics, independent of the effects of other factors also affecting fertility, such as poverty status, education and urbanization. Using a model which controls for differences in adolescent sexual activity in different areas in 1970 and 1975, the analysis found that for every 10 teenage patients enrolled in family planning clinics in 1975, about one birth was averted in 1976. Other multivariate models, which did not control for differences in sexual activity, showed changes in the same direction, though of smaller dimension. Since the family planning program averts not only births but also pregnancies that result in abortions and miscarriages, an estimate was made of the total number of pregnancies averted by the program. Based on the proportion of unintended pregnancies among adolescents that resulted in live births in 1976 (36 percent), it was estimated that for every 10 teen patients enrolled in 1975, almost three pregnancies were averted in the following year. Over the 1970s, an estimated 2.6 million unintended adolescent pregnancies were averted by the program--944,000 births, 1,376,000 abortions and 326,000 miscarriages. In 1979 alone, an estimated 417,000 unintended pregnancies were prevented by the program.

PIP: During the 1970s there was a decline in adolescent childbearing in the U.S. and, among teenagers who were sexually active, there was a decline in pregnancy rates as well. Areal multivariate analysis reveals that adolescent birthrates were reduced between 1970 and 1975 as the result of enrollment by teenagers in family planning clinics, independent of the effects of other factors also affecting fertility, such as poverty status, education and urbanization. Using a model which controls for differences in adolescent sexual activity in different areas in 1970 and 1975, the analysis found that for every 10 teenage patients enrolled in family planning clinics in 1975, about 1 birth was averted in 1976. Other multivariate models, which did not control for differences in sexual activity, showed changes in the same direction, though of smaller dimension. Since the family planning program averts not only births but also pregnancies that result in abortions and miscarriages, an estimate was made of the total number of pregnancies averted by the program. Based on the proportion of unintended pregnancies among adolescents that resulted in live births in 1976 (36%), it was estimated that for every 10 teenage patients enrolled in 1975, almost 3 pregnancies were averted in the following year. Over the 1970s, an estimated 2.6 million unintended adolescent pregnancies were averted by the program--944,000 births, 1,376,000 abortions, and 326,000 miscarriages. In 1979 alone, an estimated 417,000 unintended pregnancies were prevented by the program.

Transitions in the effect of family size on female education attainment: the case of Taiwan.

PIP: The effect of family size on female educational attainment in Taiwan during a critical period of its development is examined. A sample group of women in 1973 (ages 20-39) with varying socioeconomic backgrounds and family structures is considered. Although the father's occupational status and education are positively related to the daughter's educational level, it would appear that structural aspects of the family and number of siblings are not related. The results of this study thus reveal that family size was not related to female educational attainment with the exception of certain sectors of Taiwanese society. An explanation of these findings suggests that in the early stages of development, children's education is not as important as it later becomes when greater value is attributed to schooling as a vehicle for social mobility. At that later level of societal development, parents consider making resource allocation decisions between expenditures on education and other desired goods (e.g. children) and family size begins to have a negative affect on individual educational attainment.^ieng

Reweighting DHS data to serve multiple perspectives.

Information about health and family planning infrastructures is collected through the service availability module (SAM), an important feature of the Demographic and Health Surveys (DHS) conducted in many developing countries. The DHS samples were designed to provide a representative sample of households and women of reproductive age. Using the weights routinely provided with DHS data sets, service accessibility can be described straightforwardly at the individual and household levels. However, without further adjustment, SAM data do not provide a representative picture of service delivery at the community, or primary sampling unit, level, where the data are collected. This report proposes a methodology for reweighting the SAM data, using rural data from the Egypt DHS as an illustration, so that available family planning facilities at this level may be usefully characterized at little additional cost.

Do intentions predict fertility? The experience in Taiwan, 1967-74.

Are responses to survey questions about desire for additional children useful in predicting future fertility? Data from Taiwan covering the years 1967--74, a period of rapid fertility decline and increasing contraceptive practice, indicate that reproductive intentions and contraceptive use were very good predictors of subsequent fertility. Whether more children were wanted was itself the most important determinant of contraceptive use. Aggregate consistency between attitudes and behavior was higher than found in US surveys during the same period. Similar studies are needed to determine whether statements of intentions have predictive value in other settings as well.

PIP: 2,055 cases are studied over a 7 year period with data from a 1967 longitudinal survey of a probability sample of women aged 18-39, from reinterviews in 1970 and from the Taiwan population register, 1971-1974. Results indicate that whether a woman had a birth was strongly related to whether she wanted more children and whether she was practicing contraception. Together these 2 variables predicted whether a child was born better than any combination of other demographic or socioeconomic variables. However multivariate analysis that included socio-demographic factors, marriage duration, parity, number of sons, education, as well as desire for children and contraceptive use reveal that marriage duration was the single most important determinant of whether an additional birth occurred, with desire for children showing a somewhat lower effect. Contraceptive use and parity are about equal in importance, but the number of sons does not have a significant effect. The multivariate model also indicates that whether more children are wanted is the most important determinant of contraceptive use, with education and marriage duration much lower in importance. Marriage duration is the most important determinant of whether an additional child is wanted, but parity and number of sons have substantial effects. It is concluded that the results indicate a high degree of rationalization of the fertility process in Taiwan.