Determinants of Fertility Intentions among Married Couples in 35 Countries

Stan Becker, Johns Hopkins University
Agbessi Amouzou, Johns Hopkins University

Longitudinal studies in several countries have shown the importance of both husbands' and wives' fertility preferences for predicting subsequent fertility. This paper examines determinants of couples' joint fertility desires from cross-sectional data. We utilize demographic and health survey (DHS) data for monogamously married couples available from 35 developing country surveys (25 from Africa, 5 from Asia and 5 from Latin America). Polygamous couples were excluded because the husband's response on preferences could not be linked with data for any particular wife. Four groups of couples are defined: both spouses say they want another child; both do not want another child; she wants another child, he does not; and he wants another child, she does not. The percentage of couples in which both spouses say they want no more children varies widely from a low of 4% in Niger and Chad to a high of 70% in Brazil (median = 19%). The percentage of couples in which the spouses have discordant desires varies from a low of 9% in Niger to a high of 30% in Haiti (median = 19%). In all but three surveys (Comores, Ivory Coast and Malawi) the percentage of couples with the husband wanting more children and the wife wanting no more was higher than the percentage of couples with the opposite combination. Logistic regression with standard demographic and socio-economic covariates available in all surveys were performed on three sequential subsets of couples for each survey to assess significant determinants of group membership of couples. The covariates were: age of the wife and difference of spouses' ages, number of living children, education of the wife and of the husband measured in years of schooling completed, duration of marriage, place of residence, and ownership of a radio in the household. The first dichotomy is both spouses want more children versus all other couples. The second is both spouses want no more children, versus one spouse wants no more. The third is he wants more, she does not versus she wants more, he does not. Regression was not done if there were fewer than 100 couples for any analysis. Squared terms for age, duration of marriage and years of education were included in all regressions after likelihood ratio tests showed that they added significantly to model fit in a number of country regressions. Standard adjustments for the DHS survey designs were made. Number of living children is the variable most often significantly related to group membership, having coefficients that were significant (p<.10) in 70 of 92 regressions. Husband's education, urban residence and an index variable for the wife being older than the husband were significant in nearly equal numbers of the regressions. Interestingly, wife's education was significant considerably less than husband's education (18 significant coefficients vs. 28 significant coefficients in the 92 regressions) signaling the importance of husband's characteristics in determining couple fertility preferences. Significant coefficients in each of the three sets of regressions for the indicator variable of wife being older than the husband also provide evidence for wives' deference to husbands' preferences in the couple. For example, in the regressions distinguishing couples who both want more children from other couples, the 12 significant coefficients for this indicator variable are positive, suggesting that despite her being older, which normally increases women's desire to end childbearing, she tends to agree with her husband to have more children. Though the evidence is indirect, these are probably indications of male dominance in fertility decision-making. Fertility control programs need to understand the fertility preferences of both spouses together.