the country where almost all Sri Lankan Tamils live. Therefore when the Tamil/Hindu group is considered in the SLOPS reinterviewed subsample or its foliow-up, the SLCS, it is largely representative of Indian Tamils (see Table 2.4), who are concentrated in the tea estate sector of Sri Lanka.
with pay, and from improvement in their economic circumstances and nutrition which may have influenced their reproductive ideals positively.
Women whose husbands were in the non-primary sector also showed smaller ideal family sizes than women whose husbands were in the primary sector. Among work status categories, except for women in the oldest age group in 1982 and in 1985, those who were not working had smaller family size ideals than women working either at home or away from home. Since majority of female workers in Sri Lanka are engaged in the agricultural sector they probably prefer larger family sizes in order to have enough child labour to support their activities. Contraception was clearly related to family size ideals; those who were using no method recorded the highest ideals in all age groups considered in both 1982 and 1985.
Perhaps the principal importance of comparing of ideal family size reports in 1982 and 1985 is to identify which socio-economic categories of women tend to prefer larger family sizes than others. A largely similar pattem of socio-economic differentials in family size ideals was observed from 1982 to 1985. The overall conclusion is that, although the questions on ideal family size adopted in the two surveys were different, the reproductive ideals of young women, particularly, expressed through them approximate each other. Undoubtedly overall family size preferences in Sri Lanka are already at low levels. Mean (personal) ideal family sizes of 3.4 and 3.1 were recorded in 1985 and 1987, while the 1982 SLCPS generalized mean was 2.9 children; the estimates were not much different from each other. Since the late 1950s an increasingly large proportion of Sri Lankan women have been able to limit their family sizes to small numbers (section 1.4.5). Family size preferences sought via personal and generalized types of questions showed little difference at the aggregate level.
Moreover, at the individual level about 43 per cent of the respondents gave identical ideal family sizes in 1982 and in the 1985 follow-up survey (Table A3.7)45 Owing to 92
15 The evidence on test-retest reliability of the ideal number of children in other developing countries is found to be weak, and a large proportion of women presumably do not have fixed ideal family sizes. For instance, in Indonesia (MacDonald et al. 1978:25) only 54 per cent of respondents gave the same number of children to the identical question ‘If you could choose exactly the number o f children to have in your whole life, how many children would that be?’ (see Chapter 1).
the incompatibility of the two questions a detailed analysis of changes in ideal family size at the individual level between 1982 and 1985 was not attempted. However, in order to understand reproductive preferences and their correlates in more detail Chapter 4 examines another preference indicator, desire for more children, a measure which could not be affected by rationalization.
3.6 Summary
Since inter-item consistency analysis of the preference data from Sri Lanka shows high consistency, the theory of random answering has to be rejected. The preferences of the women being examined were held with strong intensity and high conviction. Reproductive preferences of women in the SLCPS full sample, re-interviewed subsample and 1985 follow-up sample all show high agreement. In all three samples about 90 per cent of women who wanted more children also reported ideal family sizes in excess of their actual family sizes. Consistency was also high when responses to the ‘whether last pregnancy wanted’ question were related to the variable comparing actual and ideal family size. About 80 per cent of women who reported that their last pregnancy was wanted had ideal family sizes equal to or above their actual family sizes.
The observed trend in mean ideal family size in Sri Lanka shows a considerable decline over 12 years from 3.7 children in 1975 to 3.1 children in 1987. The mean of 2.9 children obtained from the 1982 SLCPS was lower than either of these figures, but this was a result of the nature of the question used in that survey being somewhat hypothetical for older women. Most importantly, ideal family size estimates from the 1982-85 longitudinal data show little change among the young women. These days women in Sri Lanka want rather small families, although very few wish to remain childless or to have only one child. The overwhelming majority nominate two or three children as ideal, and few want larger families. Actual family size, but not age, influences reproductive ideals; when actual family size is held constant the difference between age groups in mean ideal family size becomes negligible. Although it could not be proven whether rationalization was taking place, with women having tended to
94 nominate the number of living children they had as ideal, the argument could not be ruled out for the 1985 reports of ideal family size given a clear positive association between this variable and actual family size.
Mean wanted family size and its refined version of 3.5 and 3.3 children respectively were higher than mean ideal family size in 1982. This was inevitable, since respondents wanting no more children were assumed to have wanted the number they actually had,
which in reality may have exceeded the number wanted. However, the crucial
conclusion of the present chapter is that there is strong evidence that fertility would fall remarkably, over a short period of time, if women were to implement reproductive preferences efficiently. In the 1982 SLCPS over one-quarter of women labelled then- last pregnancies as not wanted.
Differentials in ideal family size by socio-economic levels in the 1982 and 1985 data show almost the same pattern. After actual family size had been controlled for, respondents’ education levels showed the expected inverse relationship with ideal family size, while Muslims had the highest ideals of all the ethno-religious groups. Although differences appeared to be narrowing, urban and estate women favoured smaller families than rural women. Non-working women aged less than 35 years in 1982 and the equivalent cohorts in 1985 nominated lower ideal family sizes than women working either at home or away from home. Non-users of contraception in all age cohorts at both dates tended to prefer larger family sizes than users of any form of contraception. Among the youngest cohort women (less than 25 years of age) in 1982, sterilized women had the smallest ideal family sizes, but in 1985 it was the non permanent method users in the same cohort whose ideals were lowest. Within the same cohort, women whose husbands were in the non-primary sector also showed a preference for smaller families than those whose husbands were in the primary sector.