Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
- 6 Changes in Reproductive Patterns Previous chapters of this report focused on the consequences of different aspects of reproductive patterns for the health of women and children as individu- als. We examined the relationships between maternal age, birth spacing, and birth order and the health of individual women and children. The health of women, we concluded, can be improved by reducing the number of births and especially by reducing high-risk births, specifically those of high-parity and to very young and older women. The health of individual children can also be improved by reducing high-risk births, specifically births to very young women and births occurring less than 24 months after a previous birth. In this chapter we change the focus of our attention and examine the degree to which reproductive patterns vary among developing countries and the extent to which these patterns have changed as societies move through demographic transition. This chapter reviews evidence of changes in the proportions of high-risk births occurring in countries as their levels of fertility decline. As fertility in a society declines, other aspects of the society's pattern of reproduction may also change. For example, women may be more likely to compress their childbearing by delaying their first birth, having births closer together, and having their last birth at a younger age. This was what happened in South Korea (Donaldson and Nichols, 1978~. Such changes in the timing, spacing, and numbers of births in a society change the proportion of high-risk births and thus affect the health of the women and children in that society. Our review of reproductive patterns in the developing world begins with an examination of changes in birth order distributions that have occurred in countries in which fertility has declined. We then examine evidence of changes in ages 76
CHANGES IN ~PRODUCTWE PATTERNS 77 when mothers begin and when they complete childbearing as fertility declines in a society. We then examine patterns and trends in birth spacing and the relationship between contraceptive use and abortion. Finally, we examine how changes In reproductive patterns affect mortality rates. Most of the evidence presented in this chapter comes from three large-scale survey research projects supported by the Agency for International Development: the World Fertility Survey (WFS), the Contraceptive Prevalence Surveys (CPS), and the Demographic and Health Surveys (DHS). Additional information comes from other cross-sectional surveys and from national vital registration systems. Unfortunately, comparable data from the same set of countries are not always available, and the countries for which data are available are not necessarily representative of the experiences of the developing world. However, the avail- able data show considerable variation among societies in reproductive patterns and allow us to illustrate the experience of many developing countries as their reproductive patterns change. The increasing availability of more than one national sample survey of fertility in many developing countries will make it easier for future studies to examine changes in reproductive patterns. At the same time, it should be recognized that one of the factors that can contribute to a fertility decline and changing reproductive patterns is the level of infant mortality, so that there are reciprocal effects. The rates for a given event in a population depend on the rates for that event in each subgroup and the relative size of that subgroup in the population. For example, the overall infant mortality rate may be thought of as resulting from the infant mortality rates applying to each birth order and the proportion of births by birth order. The emphasis in this chapter, however, is on the distribution of births by birth order, mother's age, and birth interval, and how these are likely to change as fertility declines. In order to gauge how the infant mortality rate will change with changing reproductive patterns, it is necessary to know the level of infant mortality associ- ated with each subgroup. At the same time, it should be recognized that these subgroup rates may themselves change as reproductive patterns vary. The vari- ations in infant morality rates associated with different levels of subgroup char- acteristics provide some indication of the changes that might result from changing reproductive patterns. At the same time, it should be noted that the exact amount of change will also depend on the trends in the group-specific infant mortality rates. The emphasis in this chapter is on the distributions of births in populations, and not on the probability of births to individuals or subgroups or how these distnbu- tions are likely to change as fertility declines. For example, a decline in births to older women may mean that an increased proportion of all births will occur to younger women, even if birth rates decline for both young and older women. It is virtually certain, for example, that lower fertility will result in a decrease in the proportion of higher-order births and a corresponding increase in the proportion
78 CO=~CE~ION AND REPRODUCTION of first births. Lower fertility probably will lead to a decrease in the proportion of births to older women, but it is not clear what effect lower fertility will have on the proportion of births to very young women. There are limited grounds on which a prediction can be made about possible changes in the distribution of birth inter- vals. Several scenarios are possible. For example, if there is a reduction in breastfeeding without a corresponding increase in contraceptive use as fertility declines, birth intervals will become shorter. Average interval length could also decrease as a result of changes in the birth order distribution. Intervals between higher-order births are often longer than between lower-order births, so fertility declines that reduce higher-order births and concentrate births among lower orders could result in shorter average interval lengths. However, if contraceptive use for spacing births increases without a corresponding decrease in breastfeed- ing, the length of intervals may increase as fertility declines. These changes may occur simultaneously, adding to the difficulty of predicting the effect of changing fertility on the length of birth intervals. However, since potential changes in the distributions of high-risk births may significantly affect the health of women and children in a society, these changes need to be understood. FERTILITY DECLINES AND BIRTH ORDER DISTRIBUTION Total fertility rates in the developing world have declined from an estimated average of 6.1 during the first half of the 1950s to an estimated average of 4.1 during the first half of the 1980s (United Nations, 1988c). Significant declines in total fertility rates have occurred throughout Latin America and Asia, as is well known. However, declines in total fertility rates have not taken place throughout the entire developing world: for example, although the fertility decline in China has been extraordinary, the total fertility rates in su~Saharan African countries are estimated by the United Nations (1988c) to have increased slightly. As the total fertility rate of a population declines, the proportion of all births that are first births increases and the proportion of higher-order births (fifth and higher births) decreases. This is illustrated in Table 6.1, which shows changes in the distribution of first and higher-order births for 11 countries that have experi- enced significant fertility declines and that the United Nations judges to have reasonably complete vital registration data. In each case, the proportion of all births that are of order 1 increased substan- tially, more than doubling in some countries, and the proportion of higher-order births decreased. For example, in Malaysia the proportion of first births increased from 12 to 26 percent, and the proportion of fifth- and higher-order births decreased from 41 to 22 percent. These changes decrease the number of high- risk, higher-order births and increase the proportion of high-risk first births. We will return to the effects such distributional changes have on measures of health and mortality at the end of the chapter.
CHANGES IN REPRODUCTrVE PATTERNS 79 TABLE 6.1 Change in the Order Distnbunon of Births in Selected Countnes Over the Course of Fertility Declines by Percentage Decline in Total Fertility Rates. Proportion of All Proportion of all Births of Order 1 Bits of Order 5+ Percent 1960s 1970s- 1960s 1970s- decline Country 1980 1980 in TO Singapore .23 .44 .33 .02 65 Hong Kong .25 .43 .23 .04 64 Barbados .22 .40 .35 .10 54 Mauritius .18 .36 .36 .11 52 Costa Rica .18 .32 .45 .17 50 Chile .25 .41 .31 .09 49 Trinidad and Tobago .19 .32 .37 .19 43 Pueno Rico .27 .32 .27 .10 42 Panama .21 .29 .35 .22 42 Malaysia .12 .26 .41 .22 42 Fiji .23 .35 .36 .13 41 Source: United Nations Demographic Yearbook, venous years. MATERNAL AGES AT CHILDBEARING Because very young and older maternal ages are associated with increased risk for both women and children, we are interested in the changes that are likely to take place as fertility declines in the ages at which women begin and end child- beanng. There is considerable information on the age at which women begin childbearing. Information on age at completion of childbearing is more limited, because measuring this requires that cohorts have completed childbearing. How- ever, we can show changes in the proportion of births to women over age 35 in some developing countries that have experienced declines in fertility. Age at first birth has been rising in many developing countries. Trussell and Reinis (1989) have estimated mean ages at first birth by age groups for the 40 developing countries participating in the WFS. It is difficult to be certain that these data reflect change over time because of the tendency for women, especially older women, not to report their first birth if it did not survive. The Trussell and Reinis estimates, which are presented in Table 6.2, show increases in age at first birth in most of the countries that have experienced declines in total fertility. By comparing the average at first birth of the youngest with that of the oldest groups, we can estimate the extent to which change has taken place. For example, in
80 CO=~CE~ION AD REPRODUCTION TABLE 6.2 Mean Age at First Birth, by Age Country Age at Time of Survey 20-24 25-29 30-34 35-39 40-44 4549 Africa Benin 21.0 20.5 20.4 20.4 21.1 21.6 Cameroon 19.6 20.1 19.9 20.6 20.7 -- 22.0 Egypt 21.7 22.2 20.5 20.0 20.0 19.9 Ghana 20.4 20.7 20.6 20.5 21.2 21.2 Ivory Coast 19.0 19.4 19.2 19.9 20.1 20.4 Kenya 19.9 19.6 19.2 19.7 20.0 21.3 Lesotho 20.9 21.2 21.4 21.6 21.9 21.5 Mauritania 20.9 19.6 18.8 20.0 20.7 21.1 Morocco 22.5 21.4 20.4 19.5 19.9 19.5 Nigeria 20.0 20.2 20.2 20.9 22.5 22.9 Senegal 19.4 19.4 18.3 18.5 18.7 19.4 Sudan Worth) 20.7 20.1 19.4 20.7 20.4 21.8 Tunisia 24.8 24.2 22.3 22.0 22.1 22.7 Asia and the Pacific Bangladesh 17.5 17.1 17.1 17.4 17.9 18.1 Fiji 23.2 22.4 20.9 20.6 20.4 20.7 Indonesia 21.0 20.5 19.8 19.6 20.1 20.7 Jordan 20.2 20.5 20.1 20.2 20.1 19.7 Korea 25.1 24.1 24.2 22.7 21.8 20.7 Malaysia 24.1 23.5 22.0 20.9 20.8 20.4 Nepal 21.0 20.8 20.8 21.5 21.7 22.0 Pakistan 20.1 20.6 19.9 19.8 19.1 19.4 Philippines 23.4 23.1 23.1 22.5 22.4 23.0 Sri Lanka 25.8 25.6 22.9 22.8 21.6 21.6 Syria 22.6 21.8 21.1 22.0 22.0 22.0 Thailand 22.9 23.0 22.2 22.7 22.4 22.4 Turkey 21.8 20.9 20.6 19.9 20.4 20.8 Yemen Arabic Republic 19.9 19.9 20.9 21.4 22.3 24.1 Caribbean and Latin America Colombia 22.6 22.4 21.6 21.7 22.0 22.5 Costa Rica 22.4 22.9 21.9 21.7 21.9 22.B Dominican Republic 21.2 20.3 20.7 20.2 20.3 21.3 Ecuador 22.7 22.4 21.5 21.3 21.3 22.4 Guyana 21.6 21.1 20.4 20.4 20.1 20.7 Haiti 24.9 23.8 23.5 23.4 22.0 24.1 Jamaica 19.5 20.5 19.8 20.3 21.5 21.7 Mexico 21.8 21.8 21.3 21.2 21.1 21.5 Panama 22.7 22.1 21.2 21.2 21.1 21.2 Paraguay 23.1 22.5 22.3 22.4 21.4 22.0 Peru 22.8 22.1- 22.0 21.4 21.5 21.9 Trinidad and Tobago 22.6 23.1 22.3 21.6 20.7 20.8 Venezuela 22.4 22.2 21.8 21.3 21.4 n.a. Source: Trussell and Reinis (1989). Mean ages are based on data derived from the World Fertility Surveys and are estimated based on statistical models developed by Coale and McNeil (1972) and Rodriguez and Trussell (1980).
CHANGES iN PRODUCTS PA=E~S ~ ~ Malaysia women ages 4549 were on average 20.4 when they had their first birth. Trussell and Reinis's estimates show than in Africa, only Egypt, Morocco, and Tunisia experienced a rise in the mean age at first birth. In Asia and the Pacific, one sees more increases in the estimated mean age at fast birth, especially in countries with significant fertility declines, such as Korea and Malaysia. In the Caribbean and Latin America, however, the mean age at first birth has changed relatively little, even though many of the countries in this region have experienced notable declines in fertility. ~~ Data on changes in age at last birth are more difficult to obtain because cohorts of women must have completed their fertility in order for age at Be last birth to be measured accurately. McDonald (1984) estimated the mean age at last birth for 30 countries for women who were ages 4049 at the time the data were collected. Mean ages at last birth for these women ranged from 31.5 for Indians in Guyana to 38.4 in Kenya. Unfortunately, there are no later cohorts for comparison. Another way to see if changes have occurred in the age pattern of childbearing is to examine the proportion of all births occurring to young and older women at two different times. Table 6.3 shows proportions of births to younger women (under 20) and older women (over 35) at To time periods in 11 developing countries that have experienced significant declines in fertility. The proportion of births to women ages 35 and older declined in all these counmes. Declines were substantial in some cases from 20 to 6 percent in Hong Kong and marginal in TABLE 63 Changes in the Distribution of Births by Matemal Age in Selected Countries Over the Course of Fertility Decline by Percent Decline in Total Fertility Rates Between 1960 and 1980 Proportion of All Births Proportion of All to Women Under 20 Births to Women 35+ Percent 1960s 1970s- 1960s 1970s- Decline Country 1980 1980 in TFR Singapore .08 .04 .14 .05 65 Hong Kong .05 .()4 .20 .06 64 Barbados .21 .25 .15 .06 54 Mauritius .13 .14 .15 .07 52 Costa Rica .13 .20 .18 .09 50 Chile .12 .17 .17 .09 49 Trinidad and Tobago .17 .19 .1 1 .08 43 Puerto Rico .18 .18 .11 .07 42 Panama .18 .20 .11 .09 42 Malaysia .1 1 .07 .14 .13 42 Fiji .13 .11 .12 .08 41 Source: United Nations Demographic Yearbook, various years.
82 CO=RACE=ION ED REPRODUCTION others, especially in those countries with a relatively low proportion of births to older women in the earlier time period, such as Panama Those countries with the greatest fertility declines all experienced a substantial drop in the proportions of births to women ages 35 and older. The proportion of births to women younger than age 20 increased in 6 of the 11 populations as their fertility declined. Delaying age at first birth can benefit the health of women and their children, particularly for women who might begin childbearing at very young ages. The data on ages at which women begin childbearing reviewed above indicate that, in a number of countries, fertility declines have also been accompanied by increas- ing ages at first birth. A decrease in the proportion of births to older women is also related to declines in fertility. This change in reproductive pattern has probably benefited the health of women and children in these populations. SPACING OF BIRTHS As shown in Chapter 5, births following a previous birth by less than 24 months are associated with increased risk in many developing countries, in developed countries, and in high-mortality historical populations. Although scientists have not yet specified the precise mechanisms involved in these rela- tionships, the association is so widespread that it is prudent to be concerned about the potential effects on infant health of short birth intervals. There is limited evidence regarding the relationship between changes in the proportion of short birth intervals (less than two years) in a population and changes in levels of fertility. Table 6.4 presents Hobcraft's (1987) estimates for 18 countries of the proportion of second- and higher-order births that followed the previous birth by less than 2 years during the 10 years preceding the WFS survey. This table presents a cross-sectional view of countries at different stages in the demographic transition. Kenya and Jordan, the counmes with the highest total fertility rates, have a high proportion of short birth intervals. However, among the other countries, those with lower levels of fertility tend to have higher proportions of short birth intervals than those with higher levels of fertility. Over half the births in Costa Rica occurred less than two years after the previous birth, and their total fertility rate of 3.8 is relatively low. In comparison, 37 percent of births in Bangladesh followed within two years of a previous birth, while the total fertility rate was 6.1. If an increase in short intervals occurs when fertility declines, then there are reasons to be concerned about the increased risk to infants associated with increases in the proportion of short birth intervals. However, this cross-sectional perspective may be an incomplete or inaccurate view of the relationship between changes in birth spacing and fertility declines. Published data on changes in the proportion of short birth intervals in popula- tions experiencing declines in fertility are surprisingly limited. Table 6.5 presents data from 10 countries in which both birth interval data and total fertility rates are
CHANCES IN PRODUCTS PA=E~S 83 TABLE 6.4 Proportion of Births of Orders Two and Above Chat Occurred Less Than Two Years After the Previous Birth and Total Fertility Rates for the Five-Year Penod Prior to the Survey Proportion with preceding Country (Survey year) Interval < 2 Years Total Futility Rate Kenya (1977-1978) .40 8.2 Jordan (1976) .54 7.8 Senegal (1978) .21 7.1 Camerom (1978) .32 6.3 Ivory Coast (1980-1981) .29 6.3 Bangladesh (1975-1976) .37 6.1 Morocco (1980) .42 5.9 Lesotho (1977) .20 5.6 Peru (1977-1978) .44 5.5 Jamaica (1975-1976) .48 5.0 Colombia (1976) .53 4.6 Malaysia (1974) .41 4.6 Thailand (1975) .37 4.5 Panama (1975-1976) .47 4.4 Korea (1974) .22 4.2 Costa Rica (1976) .53 3.8 Sn Lanlca (1975) .35 3.7 Trinidad and Tobago (1977) .49 3.2 Sources: For intervals, Hobcraft (1987); for fertility rates, Goldman, Rutstein, and Singh (1985). available for some point in the 1970s and some point in He 1980s. For two of these countries, Thailand and Taiwan, we also have data from the 1960s. In ~ of these 10 populations, the proportion of birth intervals of less than two years decreased as fertility levels dropped. These data are admittedly limited, but they do present a picture opposite from that given by cross-sectional data showing that the proportion of intervals that are short is negatively associated with total fertility rates over time. If birth intervals do lengthen as fertility levels decline, improve- ments in maternal and child health should occur. However, the increase in the proportion of short intervals in Senegal should be of concern to health officials and policy makers, particularly if this reflects emerging patterns in Africa. As data collected at more than one time become available for a larger number of countries, the relationship should become clearer. Breastfeeding and postpartum abstinence are the primary traditional factors contributing to long interbirth intervals. Intensive unsupplemented breastieeding postpones the return of ovulation (McNeilly, 1977), thus extending the length of birth intervals. Postpartum abstinence, which can have mean durations Heater
84 CONTRACTION AND REPRODUCTION TABLE 6.5 Changes in the Distribution of Births by Length of Preceding Interval and Total Fertility Rates (I~FR) in Selected Countries 1960s Proportion of intervals less than 24 months TF~ 1970s Proportion of intervals less than 24 maths TFR 1980s Proportion of intervals less than 24 months TFR Senegal 21 7.1 34 6.68 Ecuador 46 5.3 28 4.38 Peru 41 5.5 26 4.1d Mexico 49 6.1 28 3.8c Dominican Republic 54 5.7 25 3.7d Colombia 53 4.6 21 3.38 Trinidad and Tobago 49 3.2 17 3.18 Sri Lanka 35 3.7 19 2.78 Taiwan 37a 5.1b 49a 3.4c 53a 2.6f Thailand 3oa 6.1 25a S.la 19~ 2.48 a Closed intervals occurring within 12 months of survey. All other birth intervals are for the 10 years prior to the surveys. b 1963-1965. c 1971-1973. d 3 years prior to survey. c Year prior to survey. f 1978-1980. g 5 years preceding survey. Sources: Senegal, Ecuador, Peru, Mexico, Dominican Republic, Colombia, Trinidad and Tobago, Sri Lanka: 1980s data provided by the Demographic and Health Survey project; 1970s data from the World Fertility Survey (A. Pebley). Taiwan: Data provided by A. HenT~alin. Thailand: John Knodel and Kua Wongboonsin, "Birth Interval and Birth Order Distributions over Ihailand's Fertility Tran- sinon." August, 1988. 1980 TI;Ks from DHS Newsletter except Taiwan. than a year in some populations, is also an important influence on birth interval lengths in some countries. Table 6.6 presents estimates of mean durations of breastEeeding and postpartum abstinence by Singh and Ferry (1984) for 20 WFS countries. Mean durations of full (unsupplemented) breastfeeding range from 2.2 months in Kenya to 7.9 months in Mauritania. Mean duration of total breastieed- ing is as high as 26.5 months in Bangladesh. Countries with longer intervals between births tend to be the countries in which durations of breastfeeding are longer. There is a wide range in periods of postpartum abstinence, with mean durations of more than a year in several countries. In most developing countnes, more women have breastfed than have used modem contraception; therefore declines in breastfeeding could greatly increase the proportion of short birth intervals (DaVanzo and Starbird, 1989). Accord- ingly, some experts have been concerned that breastfeeding in the developing
CHANCES iN PRODUCTS PA"E~S 85 TABLE 6.6 BreastEeeding and Postpartum Abstinence in 20 Counties Mean Duration (months) % of Mothers Unsupplemented Postpartum Who Ever County Breas~eeding Breastteeding Abstinence Breastfeed Bangladesh 26.5 n.a. 3.0 98 Benin 19.2 2.6 15.5 97 Lesotho 19.1 2.5 15.0 95 Ghana 17.9 4.5 10.0 92 Senegal 17.7 4.9 n.a. 98 Cameroon 17.5 5.1 13.9 98 Ivory Coast 17.5 5.0 13.1 98 Kenya 16.9 2.2 2.9 98 Egypt 16.3 7.4 n.a. 95 Sudan (North) 15.8 5.6 2.6 98 Mauntania 15.6 7.9 n.a. 98 Haih 15.3 n.a. 6.5 96 Morocco 14.2 5.5 n.a. 94 Tunisia 14.0 6.2 1.6 95 Philippines 12.6 3.3 2.8 86 Syria 11.2 5.5 1.2 95 Paraguay 10.9 2.9 1.5 93 Yemen A.R. 10.6 4.5 2.8 92 Fiji 9.4 n.a. 5.1 87 Costa Rica 5.0 n.a. 1.3 75 Source: Singh and Feny (1984:Table 5). world has declined. Millman (1986), however, found variability in trends in breastfeeding practices among the populations she examined, with decreases in some populations, increases in some, and stability in patterns in others. She concludes that there is no evidence of systematic declines in levels of breastfeed- ing, although substantial declines have occurred in some countries, for example, Taiwan. However, data concerning breastfeeding trends are limited, and it is difficult to be certain that rapid changes in breastfeeding practice are not occur- ring. Survey data from many countries indicate Tat many women desire to space their births. In Africa, spacing children has been a dominant reason for using contraception, although prevalence of modern contraception remains extremely low (London et al., 1985~. In 17 Latin American and Caribbean countries for which WFS data are available, an average of 38 percent of women con~acepting were doing so to delay their next birth, compared with 48 percent who were using contraception to prevent further childbearing (Cleland and Rutstein, 1986~. Cross-
86 CO=RACEPIION AND REPRODUCTION national analyses of contraceptive use for spacing are difficult because no stan- dard set of women has been asked questions about contraception for spacing, and the form of the questions concerning contraception for spacing has varied. While a significant proportion of women apparently wish to space their births, there is a negative association between the two primary mechanisms affecting spacing, breastfeeding, and contraceptive use. Women who breastfeed are less likely than others to use contraception, and women who practice contraception are less likely to breastfeed (Millman, 1985; DaVanzo and Starbird, 1989~. Pebley et al. (1985) found that the proportion of breastfeeding women using oral contracep- tives was generally low, but not inconsequential. Gomez de Leon and Potter (1989) present evidence suggesting that many women may either perceive breastfeeding and contraceptive use to be incompatible or they substitute one for the other. Births within 24 months of a previous birth are at increasing risk, but no clear picture emerges on the association between declining fertility and changes in the proportion of short birth intervals. The proportion of short intervals decreased as fertility declined in the six Latin American countries in Table 6.5, as was also the case in Sri Lanka and Thailand. However, the proportion of short intervals increased in Senegal and Taiwan. Trends in breastfeeding and contraceptive use for spacing are also not clear, but there is an inverse association between the two behaviors that influence spacing. CONTRACEPTIVE USE AND INDUCED ABORTION Abortion may be used when women wish to control fertility but do not practice contraception or when they experience a contraceptive failure. However, because of lack of data, it is impossible to estimate the incidence of abortion or the relationship between changes in contraceptive use and abortion in developing countries. Three patterns of abortion and contraceptive use can be identified from the limited data available for countries moving from high to lower levels of fertility (Rogers, 1988~. Abortion rates may increase during the early stages of the fertility transition, as women began to control their fertility and contraceptive use increases. This has been the experience in China, Singapore, and Taiwan, al- though a different pattern may take place in different cultural contexts. The second pattern was found in many countries of Western Europe and Japan, as the progression from high to low fertility combined with the assimilation of effective contraception was accompanied by a decline in abortion. A third pattern is characteristic of low-fertility countries in which the use of effective contraception is widespread, abortion is legal and available, and its use appears to be low. This is the current situation in England and Wales and the Netherlands. Overall then, the evidence suggests that the increased use of effective contraception can reduce the incidence of induced abortion, unless it is already low.
CHANGES IN REPRODUCTIVE PATTERNS 87 EFFECTS OF CHANGES IN REPRODUCTIVE PATTERNS ON MORTALITY RATES
88 CO=RACE~ION ^D REPRODUCTION A related issue concerns the potential for change in mortality rates caused by a change in fertility. The amount of change in mortality rates will depend on how common potentially detrimental reproductive patterns are in a population. For example, birth spacing patterns and the frequency of pregnancy at very young ages vary considerably in developing countries. The proportion of intervals that are very short is small in most countries in sub-Saharan Africa and South Asia. In these regions there is not much room to reduce this proportion further, as there is, for example, in Latin America. Rather, as fertility declines, the concern of policy makers should be to prevent an increase in the proportion of birth intervals that are detrimentally short by respecting traditional practices and by promoting breastfeed- ing and contraceptive use for spacing. Childbearing at very young ages is more common in some South Asian countries than elsewhere in the world, and there- fore the scope for improving women's and children's health through delays in the age at first birth is greater in these countries. CONCLUSION As levels of fertility decline in developing countries, their patterns of repro- duction also change. In particular, women may begin bearing children at later ages and complete childbearing at earlier ages, and the spacing between births may change. These changes in timing and spacing are likely to directly affect the health of women and children in these countries in addition to the direct benefits of having fewer children. In many developing countries in which fertility has declined, age at marriage and age at first both have increased. Delaying age at first birth is likely to benefit women and children, especially in societies in which childbearing begins at very young ages. Although published evidence on changes in birth spacing is surpris- ingly scarce, cross-sectional relationships between the proportion of short birth intervals and the total fertility rates of countries indicate that spacing tends to be closer in societies with lower fertility. However, in ~ of the 10 countries in which data on spacing are available for two or more times, the proportion of short intervals declined as fertility declined. While the available data indicate that changes in birth spacing are likely to occur as fertility declines, the direction of change remains unclear. Analysis of spacing using data from the Demographic and Health Surveys currently under way will give further indication of relation- ships between birth spacing and fertility declines. Declining fertility does lead to changes in the distribution of births by birth order; most notably, the proportion of births that are first births, which are typically higher-risk births, increases. Because of the change in the birth order distribution and the different level of risk associated with different birth orders, the paradoxical situation can en se whereby aggregate measures of mortality, such as the infant mortality rate, can increase, although the mortality rate for each birth
CHANCES IN REPRODUC.T~E PANELS 89 order may decrease. In addition, as fertility declines, the number of infant deaths will decline, even if the infant mortality rate does not change. The overall health of a population may improve as fertility declines, both because of the direct and indirect effects discussed in this report, and because of factors such as improved nutrition and improved delivery of health care. Thus, the level of risk associated with first births or close birth spacing may decrease over time. As health and demographic data continue to be collected and com- pared with information from earlier surveys, the complex associations among the numerous factors influencing the health of women and children should be better understood.
