FERTILITY PATTERNS AND THEIR DETERMINANTS IN THAILAND, 1969-1979: RESULTS FROM CROSS-SECTIONAL AND LONGITUDINAL STUDIES

Building on Easterlin/Crimmins's "Supply-Demand Model", this study focuses on the determinants of fertility in Thailand from 1969 to 1979. The data bases for the study are the National Surveys: LS1 (1969/70), LS2 (1972/73), WFS (1975) and NS (1979) and community data concerning availability of and accessibility to health care and family planning service for WFS rural population.

From the panel data, the main interest is to test how the determinants of fertility control (linking motivation: Cn-Cd, and regulation cost: RC, with use) are altered when using ex ante measurements of desired family size (Cd) and RC in the two-equation model. The result shows that there is some ex post facto bias from RC, but none from Cd. However, there is no bias if ex ante measurement of RC is used instead. Further, ex ante attitudes towards family planning is also included in the analysis and proves to improve the measurement of regulation costs.

To correct for bias in the cross-sectional analysis at a given date, analyses are done that use only data for the regulating population. Also, a new RC measure availability of and accessibility to health care from the community data is tried. The results consistently show that Cn-Cd performs better than RC when related to use.

The results seem to be consistent for all four cross-sectional data sets for the analysis of the determinants of fertility using the two-equation model. That is, the expected direction of relationship between household fertility and the proximate determinants holds and most of the coefficients are significant for all four data sets. Estimated natural fertility derived from the coefficients and the means of proximate variable is the lowest for LS1, followed by LS2, WFS and NS, and thus increases over time. For the determinants of fertility control, the result shows that use varies directly with motivation and inversely with regulation costs. However, regulation cost measured by number of methods known relates more to use than motivation measured by Cn-Cd due to the ex post facto bias from RC, as noted. In addition, motivation measured by Cn-Cd performs better than the alternative measures such as "wants no more" when related to use. (Abstract shortened with permission of author.)