MARITAL FERTILITY IN SRI LANKA: AN ASSESSMENT OF WORLD FERTILITY SURVEY DATA
This study consists of an examination of information relating to marital fertility collected in the 1975 Sri Lankan Fertility Survey. The objects of this examination are threefold: to establish the reliability of birth history reporting in the survey population and its subgroups; to describe recent trends and differentials in Sri Lankan marital fertility, in light of the estimated data quality; and to compile and possibly develop means of assessing fertility data in places such as Sri Lanka--a developing nation which has recently been undergoing substantial changes.
First, data indirectly related to marital fertility, i.e., respondents' dates of birth and marriage, are examined. Next, reported maternity histories are assessed through an analysis of the consistency of the data. Finally, the Brass parity-cumulative fertility procedure, with substantial modification, is applied.
It is shown that respondents tend to report marriage dates more completely and more accurately than birth dates, indicating that duration-specific measures are generally preferable to age-specific measures. There appears to be a bias in age reporting such that a large number of women are older than claimed in the survey. Heaping of birth dates is quite severe throughout most of the population, and the pattern of heaping is most unusual.
Overall, the reliability of the reported maternity histories appears fairly high, but some noteworthy flaws have been detected. Throughout the population, there is an apparent tendency to report too many births as having occurred during the 12 months preceding interview and too few in the 5 to 9 years preceding interview. The most intriguing results are those for respondents living on estates. These women report the lowest fertility in the nation and declining levels of early marital fertility. It is shown that the decline is probably fictitious, but that estate fertility is indeed much lower than expected. There are indications that the low rates of childbearing result from very high levels of spontaneous or induced abortion. Urban women with low levels of education, along with estate residents, exhibit the most extreme misdating of children's births. Only the former exhibit maternity history errors different in kind from others.
Marital fertility decreased sharply in Sri Lanka in the years leading up to the survey. However, almost all of the decline had occurred in just a few years, since even as late as the second half of the 1960s very little fertility reduction was evident. Some groups still registered only minor decreases in marital fertility by 1975. The most marked fertility reduction was observed among well educated women and urbanites.
One section of this analysis is methodological, consisting of a discussion of some major modifications of the Brass P/F ratio procedure. These modifications, which consist of making the ratios marriage duration-specific and using varying recent birth intervals, make the procedure better suited for use in countries with changing age at marriage and fertility, such as Sri Lanka. They also increase the ratios' ability to yield information on trends and differentials in marital fertility.