Modelling Early Childbearing as a Bounded Outcome: Adolescent Fertility in Countries and Socioeconomic Strata
Francis Ayiah-Mensah
*
Department of Mathematics, Statistics and Actuarial Science, Takoradi Technical University, Sekondi-Takoradi, Ghana.
Francis Eyiah-Bediako
Department of Statistics, University of Cape Coast, Cape Coast, Ghana.
Harry Darko Bonsu
Department of Mathematics, Statistics and Actuarial Science, Takoradi Technical University, Sekondi-Takoradi, Ghana.
Emmanuel Asare Ayim
Department of Mathematics, Statistics and Actuarial Science, Takoradi Technical University, Sekondi-Takoradi, Ghana.
Luyton Asare
Department of Mathematics, Statistics and Actuarial Science, Takoradi Technical University, Sekondi-Takoradi, Ghana.
*Author to whom correspondence should be addressed.
Abstract
Childbearing at a young age remains a pressing concern for population health and broader development. Unfortunately, there has been extensive literature based on statistical models that have not properly accounted for the bounded variable and hierarchical structure of the data concerning childbearing at a young age. The purpose of this research is to investigate the trend of childbearing at a young age using harmonised cross-national survey data. The objective of this study is to measure changes in childbearing among young women over time. The data on the percentage distribution of women aged 20-24 years who have given birth to children under age 18 were modelled using a multilevel beta regression model with random effects for countries and regions. The results indicate a significant decrease in early motherhood, with an odds ratio of −1.5 per cent per year (p = 0.000). There are significantly higher prevalence rates of early childbearing in lower-income countries than in the higher-income groups (β = 0.868, p = 0.006). Compared within countries, the risk of early childbearing increases with residence in rural areas (β = 0.191, p < 0.001) and in the poorest wealth quintile (β = 0.466, p < 0.001), but the richest fifths have a strong protective effect (β = -0.891, p < 0.001). Both region-level and country-level variations are significant. The uniqueness of this study lies in addressing key statistical limitations in existing research by treating early reproduction as a bounded proportional outcome. Some of the take-away recommendations for this study include prioritising equity-oriented policies and using ratio and hierarchical-cognisant models for monitoring teen reproduction.
Keywords: Adolescent fertility, multilevel beta regression, socioeconomic inequalities, wealth quintiles, rural urban disparities