Source: United Nations (2011 b). World Population Prospects, the 2010 Revision 96. Total Fertility Rate: The Total Fertility Rate (TFR) summarizes the information given by the ASFRs. It is defined as the number of children that a woman would have over her childbearing years if, at each age, she experienced the current ASFRs. If the ASFRs are defined by single ages, it is simply the sum of all of them. If the ASFRs are defined by 5-year intervals, it is the sum of the ASFRs multiplied by 5. The TFR is a synthetic indicator (it uses women who live at this moment as members of a fictitious cohort), to compare the level of fertility over time, and among countries or within a country. In the period 1950–2010 the TFR in the world decreased from around 5 children per woman to around 2.5, with major regional differences. In Central America, Eastern Asia and Northern Africa, the TFR drastically declined (from 6.7 to 2.3, 5.6 to 1.6 and 6.9 to 2.8, respectively), whereas in some parts of Africa, especially Western and Middle Africa, the decline was relatively modest and the TFR remains above 5 children per woman. Part of the explanation is that contraceptive use in Africa was in 2007 considerably lower than elsewhere, with only 28 per cent of women of reproductive age who were married or in union using any method (United Nations, 2011 a).
97. The TFR is different from the life time fertility of a cohort, which represents the average number of children of all the women born in a given year (who constitute a cohort) and having survived until the age of 50 at least. In theory, this figure can be directly estimated from census data from the question on number of children ever born alive. Of course, this indicator can be calculated only for generations having completed their reproductive life, that is to say women aged 50 or over. Yet older women often underestimate the number of children that they have had during their life time, as they tend to omit children that died while still very young. In the more developed countries, there has been a major divergence between TFRs and the life time fertility of women in recent decades as younger women, who were increasingly active in the labour market, postponed childbearing until they were in their thirties. The short term effect of this was to depress the fertility of younger women, while the higher fertility of women in their thirties was to materialize only later. Consequently, Europe passed through a period of extremely low TFRs in the 1990s, even though the life time fertility of women later turned out to have changed much less.
98. General Fertility Rate: Closely related to the Total Fertility Rate is the General Fertility Rate (GFR). But unlike the TFR, which is an index, rather than an actual rate, the GFR is a true rate, defined as the number of births that occur during a year, divided by the average number of women of reproductive age (15-49 years) during the year. In other words, it is the proportion of women of reproductive age that will have a childbirth during a given year.
99. The Parity Progression Ratio at parity n is the proportion of women who already have n children who will go on to have n+1. It can also be refined, to reflect not only the number but also the composition of existing children (see Chapter 5). In practical terms, it is computed in the same way as the TFR, but limited to women who have had exactly n live-born children, regardless of whether they are currently alive or not. It is usually computed from DHS data. Computing it based on census data requires special care, due to the limitations of census data explained in Methodology Box 4. A reasonable approximation is to compute the ratio as described above, based on children born during the past 12 months, and then multiply the result by the same correction factor that was applied to obtain the corrected TFR from the apparent overall ASFRs (not specific by birth order) of the census. Alternatively, one may compute the percentage of women aged 45-49 years who have more than n children among those that have at least n. The disadvantage of the latter method is that it reflects the fertility experience of older women, which may not be entirely representative of current fertility. The following table from the 2000 census of Cape Verde was computed using this latter procedure.
Source: INE, Cape Verde
Given the way the computation was carried out, this means, for instance, that 70.1 per cent of the rural women in Cape Verde who had at least 8 children actually had more than 8, whereas 29.9 per cent had exactly 8.
100. Adolescent Birth Rate: The adolescent birth rate is the ASFR for women aged 15-19. Fertility levels among women in this age group are relevant to the status of women, since women who bear children early in life often forego the opportunity to study or find employment outside the home, in addition to consequences for health and human rights noted earlier. Maternal mortality increases steeply for progressively younger mothers under age 18. The proportion of school leavers among young mothers is also higher than among adolescents who do not have children and are not pregnant, even though the direction of the cause-effect relationship in this case is not clear. Because of its potential negative effect on the education and employment of young women, the adolescent birth rate is one of the indicators for the monitoring of Millennium Development Goal 5.B. It is also the only fertility-related indicator in the Minimum Set of Gender Indicators approved by the UN Statistical Commission in February of 2012 that can be computed from census data. In the more developed regions, the average adolescent birth rate in 2005-2010 was 24.0 births per 1,000 women aged 15-19. In developing countries the range of variation of the adolescent birth rate is considerably larger: from below 5 to over 200 births per 1,000 women aged 15-19. The highest rates are recorded in Niger (207.1), the Democratic Republic of the Congo (201.4), Mali (186.3), and Angola (171.1) (United Nations, 2011 b).11 101. As was mentioned in Chapter 2, an alternative indicator sometimes used for quantifying adolescent fertility is the proportion of all births that occur to adolescent women. Unless the purpose of this indicator is to plan services, to make them more adequate for the age profile of the typical client, this indicator is not recommended because it can convey seriously misleading impressions with respect to the time trend of adolescent fertility.
102. The 2010 Human Development Report of UNDP uses the adolescent birth rate as one of the five indicators of the Gender Inequality Index (GII). One drawback is that no attempt was made to measure to what extent early fatherhood affects adolescent boys and what consequences this may have for their future. While it is known that adolescent fatherhood is less common than adolescent motherhood, as many of the fathers of these children are older than 20, omitting the information makes it impossible to draw meaningful comparisons with the situation of men. The GII, therefore, presents adolescent fertility as an exclusively female problem.
103. In agrarian societies, where children are valued for their labour, family continuity and as an insurance against the risks of old age, women are valued for their ability to bear children. Hence, a woman who is unable to bear children is stigmatised as being inadequate and having failed the husband’s family and clan. Such women are very likely to end up in a polygamous union as the husband procures another wife who can bear children, or may end up divorced altogether. Either situation is not desirable for a woman, as it communicates her lower status in relation to a man’s. In the 2008 census of Malawi, infertility was used to refer to women who had not had a child by the age of 45 years. In this census, infertility was observed to be on the decline, from the 4.1 per cent recorded in 1987 to 3.6 per cent recorded in 2008. A DHS study by Rutstein and Shah (2004) used a stricter criterion, limited to women aged 40-44 who had been married for at least five years. Consequently, they found a lower figure, of 1.6 per cent. It is recommended that census analyses should at least exclude women who are not currently married or in unions. In some African countries, such as Chad, the incidence of infertility, according to the DHS criterion, can be much higher (7.3 per cent). It can also be substantially higher among some specific ethnic groups.
104. Time spent caring for dependent children. One of the major implications of high fertility is that it ties one of the adult household members, usually the mother, to the home for many years, to care for dependent children. How many years of care each child needs depends on the circumstances of each country. In countries where education is universal and starts at the pre-school level, women will often be able to take on responsibilities outside the home as soon as their children reach age 3. In other countries, where pre-school education is non-existent and primary school education deficient in its coverage, they may have to care for them full-time until the child is age 8 or 9. Apart from the number of children, the spacing between children matters. If fertility is concentrated in a relatively short age range, the time during which there are dependent children in the home is compressed. If fertility is spread out over a longer age range, women will spend most of their reproductive years caring for dependent children. Finally, time spent caring for children depends on infant and child mortality.
105. A measure of the number of years spent caring for dependent children can be computed from census data. To this end, one needs to determine for each woman whether she has children of her own, below the accepted age for the country after which they are no longer to be considered predominantly dependent on the mother. Alternatively, one may assume that children going to school are no longer (completely) dependent on their mothers. In more complex households there may be problems in establishing which children belong to each potential mother.12 Once this has been established, one determines the proportion of women at each age who are caring for (their own) dependent children. Summing these proportions over all ages yields the mean number of years that women spend caring for their dependent children. This assumes both that women are the ones caring for their dependent children and that they only care for their own children. One may refine the method by applying it not only to the mothers of children, but generally to all adult women present in the household that do not work or study, assuming that adult males, even if they do not work or study, generally do not have a major role in caring for dependent children. Except for a small number of censuses (Australia, Republic of Korea) that ask questions about this topic, there is generally no way to avoid such assumptions.
Figure 3: Scatter plot of the Total Fertility Rate as a function of the female/male ratio (x100)
Female/male ratio of population with at least secondary Female/male ratio of shares in parliament
Sources: UNDP Human Development Report 2010 and UN Population Division World Population Prospects, 2010 Revision
106. Apart from the Adolescent Fertility Rate, the Minimum Set of Gender Indicators approved by the UN Statistical Commission in February of 2012 contains three other fertility-related indicators. However, none of these can be computed from census data:
1. Contraceptive prevalence among women who are married or in a union, aged 15-49;
2. Ante-natal care coverage; and
3. Proportion of births attended by a skilled health professional.
6. Multivariate and further gender analyses 107.Caveats for fertility correlates of geographical/subnational units. If subnational estimates are available for basic gender data, such as various gender inequality measures, these can be correlated with fertility levels to show the relationship between gender inequality and fertility. Fertility can also be related to the percentage of female-headed households or women that participate in the labour force, at the level of different geographical units. This is not fundamentally different from some of the tabulations shown in the previous section, except that the objective is not only to show the difference between population groups but to propose some sort of systematic quantitative relationship between the characteristics of those groups (e.g. that groups that exhibit greater gender inequality tend to have higher fertility). There are, however, some caveats with respect to correlating different indicators across population groups:
a) Care must be taken to avoid using gender inequality measures which already contain fertility as one of their components, which would result in a tautological relationship. For this reason, the Gender Inequality Index of the 2010 Human Development Report is not recommended, as it is already partly based on the Adolescent Birth Rate. In the example below, two other indicators from the same report have been used, namely the female/male ratio of population with at least secondary education and the female/male ratio of shares in Parliament, both applied to countries, rather than subnational population groups. Note that the indicator of shares in Parliament may not be applicable to subnational analyses; it is included only for illustrative purposes.
b) The choice of the particular gender indicator is important. In the example shown here, the female/male education ratio shows a moderately strong negative relationship with the Total Fertility Rate across 143 countries (r = -0.69). The female/male ratio of Parliamentary representation has a very weak relationship with the Total Fertility Rate (r = -0.11).
c) One has to be alert to the possibility that the relationship found may be due to the fact that both indicators are perhaps determined by a third one, rather than being the result of any direct causal relationship between them. This issue will be taken up in the next sections.
108. In examining gender issues, the researcher begins with univariate and bivariate analyses to define a potential issue or problem as it may relate to a patterned relationship across women and men, girls and boys. Then, multivariate data analysis can be useful to differentiate correlations with causations, and to pinpoint what specific variable or variables may be the causes for a differential status or level of opportunity by sex. The primer box on multivariate analysis provided in Box 1 within chapter 2 of part 1 of this manual, entitled Multivariate Analysis to Disentangle Intra Group Variability and Interrelationships, may be useful to review prior to considering the following cases.
109. Bivariate analyses of statistical relationships can be misleading if they are interpreted to show a causal relationship. This is illustrated by the example above of the correlation between the female/male ratio of population with at least secondary education and the Total Fertility Rate. Although this relationship is moderately strong (r = -0.69), it is weaker than the correlation of the overall Human Development Index (HDI) with the Total Fertility Rate (-0.84). The correlation of the HDI with the female/male education ratio is not quite as strong (0.72), but nevertheless considerable. This begs the question whether the moderately strong relationship between the first two variables is not simply due to the fact that both are reflections of the general level of development of the country, as measured by the HDI, rather than to a specific relationship between the two of them. This indeed turns out to be the case. To show this, the TFR was estimated as a quadratic function of the HDI. The residuals of this HDI-based TFR with respect to the actual TFR were then plotted as a function of the female/male education ratio. The results in the graph below show that, once the association of the TFR with the HDI is removed, the residuals display an almost completely random pattern (r = -0.04). This does not mean that the basic idea that greater gender inequality is associated with higher fertility is incorrect; it just means that the association cannot be statistically demonstrated in this manner.
Figure 4: Scatter plot of the Total Fertility Rate as a function of the female/male ratio (x100) of population with at least secondary education after removing the joint association with the HDI
Sources: UNDP Human Development Report 2010 and UN Population Division World Population Prospects, 2010 Revision
110. Another illustration of the importance of using multivariate analysis, rather than simple cross-tabulations, for drawing conclusions on causal links is a study by McKinnon, Potter and Garrard-Burnett (2008) on differentials in fertility and family formation among adolescents in Rio de Janeiro. At first sight, the data from the 2000 census used in this study conveyed the impression that adolescent fertility among young women without religious affiliation was more than twice that of Catholics and that Pentecostal Protestants also had higher adolescent fertility rates than Catholics. However, interpreting this finding as an indication of different religious dispositions towards adolescent fertility would be quite misleading. As it turned out, Pentecostal Protestants also have higher rates of having lived with a spouse or partner, have proportionally more non-white members and reside in areas with lower overall mean household incomes. Therefore, the researchers used a regression model in which the probability of giving birth was a function of ever having lived with a partner, migrant status, educational level, age, race, religious composition, mean income and other indicators to characterize the relative prosperity level of the place of residence. Once all of these explanatory factors were considered, Pentecostal Protestants actually had a 23 per cent lower adolescent fertility than Catholics with similar socioeconomic characteristics. Young women without religious affiliation continued to have a higher fertility than Catholics, even with these controls, but the difference fell considerably, from more than double to only 29 per cent.
111. In the 2001 census of Nepal, 43.5 per cent of divorced women aged 15-49 had never had children. A likely explanation is that childless women had a higher than normal probability of being divorced by their husbands. However, it is not entirely impossible that the high incidence of childlessness among divorcees is a result of the fact that many divorces occur at an early age, before the woman has had an opportunity to bear a first child. The following approximate procedure provides a way to quantify by how much being childless raises the odds that a woman will be divorced. It is easiest to apply if the variables are all computed for a specific age interval, e.g. ages 35-39 (mean age of 37). Basically what the procedure consists of is the computation of the estimation of the number of divorced and childless women under the null hypothesis of total independence of these two phenomena.
To this end, one needs the proportions CL(i) of childless women by age (preferably in single years) in the general population and the equivalent proportions D(i) of women that are divorced by age i. If it is further assumed that there is no childbearing or divorce before age 15, the expected proportion of women that are both childless and divorced by age 37 will be
This can be compared with the actual proportion of women in this condition. The ratio between these proportions can be interpreted as the factor by which being childless increases the probability of being divorced.
Note that the procedure is not entirely conclusive in proving that the high percentage of childless divorcees is due to the discrimination of childless women. There are at least three other factors that need to be considered in this context:
The computation of the age-specific divorce rates is based on transversal observations and does not reflect the life experiences of true cohorts, which may bias the results.
The higher childlessness of divorced women may not reflect discrimination against the woman because of her childlessness but the fact that many of these marriages were unhappy ones in which intercourse rarely took place.
The procedure also assumes that childbearing is terminated after a divorce, which may not be entirely justified in societies where informal unions are common.
112. Other gender-relevant contextual factors that have been related to fertility include the household structure. Moultrie and Timaeus (2001), for example, studied how the household composition influences the fertility of women of twenty years or older in South Africa. For these women, they hypothesised that living arrangements mediate between their socio-economic and cultural characteristics and the number of children that they have borne. The focus was on whether women lived with a husband, or with relatives of their parents’ generation, or with relatives of their own generation. Living with relatives from the previous generation was found to have a negligible net impact on the lifetime fertility of mothers. However, women who lived with relatives from their own generation had borne about 20 per cent fewer children than other women of the same age after controlling for the impact of household income, the woman’s schooling, regional differentials and urban-rural residence. Unmarried and separated mothers had about 20 per cent fewer children than married mothers of the same age.
7. Interpretation, Policy and Advocacy 113. Two types of data that can almost never be obtained from censuses are the contraceptive prevalence rate and data on fertility preferences. Both of these require specialized interviewer training that can be provided in fertility surveys, but that would be too cumbersome for a census.13 Yet both the contraceptive prevalence rate and fertility preferences can have major gender implications. “The ability of women to control their own fertility is absolutely fundamental to women’s empowerment and equality. When a woman can plan her family, she can plan the rest of her life. When she is healthy, she can be more productive. And when her reproductive rights—including the right to decide the number, timing and spacing of her children, and to make decisions regarding reproduction free of discrimination, coercion and violence—are promoted and protected, she has freedom to participate more fully and equally in society.” (UNFPA web page on Gender Equality).
114. Conflicting preferences between husbands and wives can affect aggregate fertility outcomes (Voas, 2003). Note the following table that compares the desired family sizes of men and women in DHS surveys in different parts of the world:
Table 6: Mean ideal family sizes for men and women in the DHS Women Men Women Men
Haiti (2005-06) 3.2 3.3 Tanzania (2004-05) 5.4 5.9
Indonesia (2007) 2.8 3.0 Ukraine (2007) 2.0 2.1
Kenya (2008-09) 5.5 4.6 Zambia (2007) 5.1 5.7
Source: DHS Statcompiler
115. The large differences found in some countries, particularly in Subsaharan Africa, may be one of the reasons why fertility in these countries remains high, although the table above indicates that women themselves in Sub-Saharan Africa also have high fertility preferences. Having little property right and being treated essentially as a form of property to be exchanged for material goods between families, women in the polygamic system of sub-Saharan countries are especially vulnerable when they become spouseless or childless. Without the right to inherit the property of her husband, a wife in this system is motivated to maintain high fertility, hoping that at least one of the surviving children is a son on whose inherited field she can continue farming after her husband’s death (Boserup, 1970).
116. On the other hand, fertility preferences in some developed countries are significantly higher than actual fertility levels and the literature suggests that gender factors may play an important role in the explanation of these disparities (Sobotka, Goldstein & Jasilionine, 2009). Note also that in some countries (Bangladesh, Haiti, Indonesia, Malawi, Peru, Rwanda, Turkey, Ukraine) the differences are negligible or even reversed. Depending on the particular application of these data, it may be possible to construct proxies that work reasonably well for certain purposes. This is particularly the case if the objective is to combine information on fertility preferences from a DHS with other kinds of information that are only available or can only be disaggregated at the desired level by using census data. For some further explanation of how this may be done, see section 6.
117. In order to ensure that “all couples and individuals have the basic right to decide freely and responsibly the number and spacing of their children and to have the information, education and means to do so” (ICPD, principle 8), two main advocacy strategies have been found to be successful:
a) Tackling unwanted early pregnancy by providing reproductive health information and youth-friendly services to young people;
b) Investing in girls’ education and empowerment more generally, in order to benefit the young women themselves, their future families, their communities, and their countries. Both strategies can lead to informed reproductive decision-making and delayed childbearing.
118. Using census data in the areas of education and fertility, advocacy material may be able to show educational disparities in who has children at young ages, thus making the case for investments into girls’ education on demographic grounds. Using census and DHS data on women’s exposure to pregnancy and child birth health risks, contrasted with information on government funding for reproductive health care, advocates can lobby Ministries of Health and other government decision-makers to change their budgeting decisions to improve women’s health.