Doctoral Dissertations

Date of Award

5-1996

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Sociology

Major Professor

Donald Hastings

Committee Members

David Sylwester, Ian Rockett, Michael Benson

Abstract

World Fertility Survey (WFS) data have been used in several comparative studies to identify selected covariates of the risk of pregnancy and to describe common patterns of birth interval length in different countries (Trusell et al. 1985; Rodriguez et al. 1984). These studies have found that the effects of socioeconomic covariates on the risk of pregnancy remain even after biological covariates are included in the analysis. These results challenge the theoretical framework of proximate determinants (Rindfuss et al.1987). However, the validity of these results may be questionable because of the lack of complete information on the biological covariates. As a result of missing information, models previously developed do not adequately control the effects of biological covariates on the risk of pregnancy and hence may underestimate or overestimate the effects of socioeconomic covariates. All of these studies treat socioeconomic covariates as fixed covariates and lack detailed information on the two most important biological time-varying covariates that directly effect the risk of pregnancy--breastfeeding and contraceptive use. This dissertation investigates the effects of socioeconomic covariates on the risk of pregnancy by month. Having monthly data points permits the investigator to determine the impact of changes in both biological and socioeconomic covariates on the risk of pregnancy. It broadens understanding of birth-pregnancy interval dynamics in three ways: 1) It provides a model that quantifies the effect of biological and socioeconomic covariates on the risk of pregnancy using a calendar history of events from the Peru 1991-92 Demographic Health Survey (DHS) data. Unlike previous fertility data sets, the DHS data contain monthly information over a five year period for the following variables: breastfeeding, contraceptive use, and woman's work status allowing these covariates to be treated as time-varying. 2) It illustrates in detail how to manipulate the DHS data to create an event history file and a new way to handle time-varying covariates. 3) Finally, a set of programs were developed to facilitate the researcher's use of event history analysis. The software enables the user to manipulate data from any of the DHS data sets, to create life tables to be used as input data for the log-linear piecewise exponential hazard model, and to estimate the likelihood of different models. The final model developed includes eight main effects: the woman's age, birth order, breastfeeding, length of the previous birth interval, contraceptive use, place of residence, and employment; plus the interaction term: the woman's age and birth order. Prior investigators classified contraceptive methods into efficient or inefficient methods following the typology of the WFS. Results from this research show that this practice combines contraceptive methods with large differences in their effect on the risk of pregnancy. Thus, the effect of contraceptive use has been underestimated in previous research. The parameter estimates for the nine contraceptive methods suggest that the methods should be categorized into four groups rather than two categories efficient and inefficient. The four categories are: 1) very efficient-sterilization, 2) efficient--pill, IUD, and injections, 3) partially efficient--condoms and diaphragms, and 4) inefficient-- rhythm method and withdrawal. The results on woman's work status reveal that it does not matter if women work at home or outside home, but whether they are paid for the work they perform. Women who do not work and women who are unpaid workers have the highest risk of getting pregnant. This finding has strong implications, because women typically carry most of the burden of non remunerated production. Although this dissertation finds some support for the finding that controlling for biological covariates does not eliminate the effects of socioeconomic covariates on the risk of pregnancy, it does show that the use of precise information on time-varying biological covariates strongly increases the significance of biological covariates. As in previous studies, use of length of previous birth interval as a covariate is recommended to correct for correlation between birth intervals coming from the same woman. However, it is not the most important covariate, instead contraceptive use is the most significant covariate. This finding suggest that the birth experience of women is not dictated by past experience but rather is more dependent upon present birth control use. In short, detailed information on biological and socioeconomic covariates and the way in which we handle time-varying covariates provides better assessment of the effect of these covariates on the risk of pregnancy.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS