Fertility Assumptions

We chose to single out the one structural argument about the determinants of the future course in fertility that would feature most prominently in most of the expert interviews. The choice was not difficult, because in addition to the rather diffuse reference to all kinds of government policies, female education clearly stood out as the single most important factor mentioned. Almost all of the national experts from the different countries in Asia involved in the exercise consistently mentioned that they thought that the improving level of female education has been and will be the main driver of fertility decline.

The line of argumentation in these interviews seems clear: The combination of great educational fertility differentials in Asia (more highly-educated women have significantly lower fertility) with the fact that younger women are and will be more educated than older women, greatly contributes to fertility decline. These educational fertility differentials are pervasive all over Asia and in countries with very different overall levels of fertility.

Table 1 gives the recent educational fertility differentials for women in India and China. It shows the total fertility rate (TFR) for four categories of women: those without any formal education and those with some primary, secondary and tertiary education. The precise definition and discussion of these categories and data sources are given in Lutz & Goujon (2001). The table clearly and impressively shows how in these two major countries, higher education is associated with lower fertility. The total TFR gives the fertility of the total population with weights of the different educational groups corresponding to the current educational composition in India and China.

Table I

Education-specific differentials and total TFR for women in India and China.

84 W. LUTZ & S. SCHERBOV

Figure 1 gives the age-, sex-and education-specific population pyramid for India as estimated for the year 2000. The pyramid clearly shows that younger cohorts of women in India are better educated than older ones, although the gender gap is still significant.

Age

Population (millions) DNo education DPrimary DSecondary •Higher

Figure 1. Population pyramid for India, 2000.

When we think about the future of the national-level fertility in India, we must differentiate between two different effects: (a) the change in the educational composition of the population, and (b) the fertility trends within each educational group. We will first discuss these factors separately and then ( c) study their joint effect.

(a) Alternative trends in education

As can be inferred from Figure 1 future improvement in the educational composition of the population is a near certainty. It is already pre-programmed into the age structure. The younger, better-educated cohorts will inevitably become older and replace the older, less-educated cohorts.

Disregarding the unlikely case of massive adult education programs only the future education of the young cohorts is uncertain at this point, depending on future school enrolment rates at different levels. Policies can make a difference here. We capture this difference through two alternative and rather extreme scenarios on girls' and boys' education in India, one (pessimistic) scenario in which all enrolment rates stay constant in the future [scenario A: constant enrolment] and another in which India manages to implement the ambitious education goals as defined in the 1994 Cairo World Population Conference (ICPD) [scenario B: ICPD]. This highly optimistic scenario assumes the elimination of the gender gap in primary and secondary education by 2005-10, 90 percent net primary enrolment by 2010-15 and secondary enrolment of 75 percent by 2025-30, as well as an increase in transition to tertiary education by 5 percentage points until 2025-30. Trends between

Probabilistic Population Projections for India 85 2000 and the target year are based on linear interpolation. The results from these two scenarios are depicted in Figure 2 for the year 2030.

Figure 2 clearly shows that the two scenarios are virtually identical in 2030 for men and women above age 55. For younger cohorts the constant enrolment rates scenario shows an essentially frozen educational composition with a significant gender gap in education being maintained. Under the ICPD scenario the educational composition improves significantly for the younger cohorts and the gender gap essentially disappears below age 25.

(b) Trends in education-specific fertility

During the course of the Asian MetaCentre exercise, there were quite some discussions about what should be assumed in terms of future education-specific fertility trends. We considered three different options for dealing with this issues: ( 1) assuming proportional fertility changes in all educational categories, (2) assuming convergence of all educational fertility trends to one target level, or (3) assuming that education-specific fertilities will move to the observed levels of another country that is already further advanced in the process of fertility decline. Option (1) is not meaningful as a general rule because in some countries the fertility of university graduates is already so low that no further declines will be expected, despite declines in average fertility. Also, for countries for which time series exist, the declines do not seem to go in parallel. Option (2) is not meaningful in this context, because if we have complete convergence, by definition the changes in the educational composition do not affect aggregate fertility. And substantively all of the Asian countries studied maintained significant differentials, even at very low aggregate fertility levels. For these reasons we chose Option (3), which is consistent with the frequent demographic practice to think in terms of analogies, as is more generally done in the context of the demographic transition. Specifically, for the case of India presented here, we assumed that in 2030, India will have education-specific fertility rates comparable to those of China today.

What is the substantive justification for this particular assumed analogy between China today and India tomorrow? First, the aggregate TFR in India in 2030 is assumed by most projections to go to a level similar to that in China in 2000. Also, India and China are both huge and greatly heterogeneous countries. In this sense the national-level fertility differentials by education do not just reflect some volatile patterns as they may appear in small populations. Finally, the data given in Table 1 show declines in each educational group of between 26 and 45 percent or on average one-third. A sensitivity analysis that assumes that each of the fertility rates in India in 2000 declines linearly by one-third until 2030 yields virtually the same results as the assumed move to the Chinese pattern of 2000.

(c) Combined scenarios

Table 2 defines 12 scenarios that result from the cross-classification of different future trends in the educational structure of the population, and different trends in education-specific fertility rates. The first four scenarios that keep the educational structure of the population frozen are of a purely hypothetical nature because, as has been mentioned above, it is already embedded in the age structure that the younger, more educated age groups over time will move the age scale and improve the average education of the female population of reproductive age. But it is still important to talk about this hypothetical case of the educational composition by age remaining constant in its current form, because it serves as a point of reference in the minds of experts thinking about this issue. For this reason Figure 3 shows the aggregate level TFR resulting from scenarios 1 and 2 in 2050 as black dots to the right of the figure. If the educational composition remains frozen and the education-specific fertility remains constant (scenario 1), then clearly the aggregate TFR

86 W. LUTZ & S. SCHERBOV

INDIA,ICPD education goals scenario, year: 2030

Age Period of Birth

INDIA,Constant enrolment rate scenario, year: 2030

Age Period of Birth

Figure 2. Education-specific age pyramids for India under the two alternative education scenarios in 2030 as described in the text.

Probabilistic Population Projections for India 87 remains constant over time. Scenario 2 gives the case in which the educational composition remains frozen, but education-specific fertility rates decline as discussed below. In this case the aggregate TFR declines by about one child, which is the fertility trend effect that is completely free of the effect of the changing educational structure.

Table 2

Definition of 12 scenarios combining different possible trends in the educational structure with different assumptions about education-specific fertility trends.

Structure constant Enrolment rates ICPD goals for (purely hypothetical) constant enrolment

l 5 (c.enr-c.fert) 9 (ICPD-c.fert)

2 6 (c.enr-medium) 10 (ICPD-medium) 3 7 (c.enr-high) 11 (ICPD-high) 4 8 (c.enr-low) 12 (ICPD-low)

Figure 3 gives four lines for the cross-classification of the two assumptions for education-specific fertility (constant versus linear change to the Chinese pattern) with the two education assumptions (constant enrolment versus the ICPD scenario). When comparing scenarios 5 and 9 and 6 and 10, respectively, we see that the changing educational structure makes a significant difference, even with identical education-specific fertility trends. By 2050 this difference accounts for more than half a child, i.e., more than one-third of the level of fertility as given by scenario 10. If we also consider the hypothetical case of a frozen education structure (scenario 2), the difference due to differential educational structures becomes almost one child. This clearly illustrates that the experts have made a valid and quantitatively important point when they suggested that the changing educational structure would be a major force towards lower fertility.

4.0 ··---··-···---··--·-····-··---·-···-·-· -···· ·--·---·---·-··--·--···--···-···--···-···-··---·----····--·--···,

Figure 3. Selected scenarios for combining education assumptions with education-specific fertility trends.

88 W. LUTZ & S. SCHERBOV