Page 157 - IJPS-11-5

International Journal of
Population Studies Contraception and fertility in Zambia

of fertility dynamics in SSA (Ariho & Kabagenyi, 2020; change in fertility dynamics over time can be attributed
Bietsch et al., 2021; Phiri et al., 2023; Wasswa et al., to compositional differences between the surveys and
2021). The independent variables encompassed maternal varying effects of selected independent variables. Although
age categories (15 – 24, 25 – 34, and 35 – 49 years old), trend analyses are often conducted from longitudinal
adolescent ages (15 – 19 years old), residential setting studies, such studies are limited in many SSA countries.
classification (rural or urban), educational attainment Decomposition methods have been developed to
levels (no formal education, primary education, secondary determine the contribution of factors to outcomes using
education, or tertiary education), and marital status cross-sectional data despite some limitations.
classifications (single, married, or previously married),
household wealth index (poor, moderate, or rich), 2.4.1. Decomposition analysis
religious denomination (Catholic, Protestant, or other), The statistical tool, Blinder–Oaxaca decomposition analysis
employment status (working or not working), age at first technique, was developed by Blinder Oaxaca in 1973
sexual debut (<15 years, 15 – 19 years, 20 or above), age (Blinder, 1973) and was later generalized by Neumark in
at first marriage (<15 years, 15 – 19 years, 20 or above), 1988 (Neumark, 1988). This approach allowed the outcome
ideal number of children (0 – 3, 4 – 5, 6+) and exposure to variables between the two groups to be broken down into
media FP messages (yes, no). components explained by observed characteristics and
estimated coefficients while controlling for all confounding
2.4. Statistical analysis factors. The Blinder–Oaxaca decomposition technique has
The analysis conducted in this research encompassed both been utilized in multiple studies for the purpose of analyzing
descriptive and multivariable decomposition techniques to models with binary dependent variables (Sinning et al.,
examine the shifting patterns in contraceptive utilization, 2008). The decomposition formula to compute the mean
fertility rates, and teenage pregnancies. The first stage difference in the binary outcome (Y) between two groups
analysis involved a trend analysis of contraceptive use, or periods was decomposed as in Equation I (Powers &
fertility rate, and teenage pregnancy rate for the six ZDHS Yoshioka, 2011).
datasets. This was performed to measure trend change for ȲA−ȲB = (EβA(YiA|XiA)−EβB(YiB|XiB))+(EβA(YiB|XiB)−
the three measures over time. The second level involved EβB(YiB|XiB)) (I)
conducting a decomposition analysis to examine the effect
of contraceptive transition and sociodemographic factors In the decomposition formula, the recent ZDHS
on fertility dynamics during the period 1992 – 2018. 2018 and reference ZDHS 1992 surveys are denoted by
A and B, respectively. The component labeled “E” in the
The third level involved examining the effect formula refers to the differential attributed to variations in
of contraceptive transition and sociodemographic endowments or characteristics of women, also known as
independent factors on the decline in teenage pregnancy the explained component or compositional characteristics
rates. A two-component Blinder–Oaxaca multivariate effects. The “C” component represents the differential part
decomposition analysis was employed to assess and related to coefficient or effect variations, commonly known
quantify the impact of contraceptive transition on fertility as the unexplained component or coefficient effects. “A”
dynamics in Zambia. denotes the time being compared, while “B” serves as the
To examine the contribution of contraceptive transition reference time. Therefore, “E” represented a hypothetical
to fertility rate reduction in Zambia, Poisson regression comparison of the outcome disparity from the viewpoint
modeling was utilized. Fertility rate reduction was of the group “A.” Group “B” viewed the outcomes of
measured using trend reduction in children ever born component “C” through a counterfactual comparison
between 1992 and 2018. Given that children ever born is (de-Boer & Rodrigues, 2020; Powers & Yoshioka, 2011).
a count variable and may exhibit some dependence, the The “mvdcmp” command in Stata software (version 17) was
single-level Poisson regression method was chosen as the used to perform a multivariable decomposition analysis of
primary analytical method. This model is typically utilized trend change in fertility dynamics between the DHS data
in situations in which the count data does not exhibit any points in Zambia.
dispersion, or when the mean value of the data is equivalent Blinder–Oaxaca decomposition analysis is unique in its
to the variance. ability to isolate and quantify the specific contribution of

A multivariable decomposition analysis for the increased contraceptive use to changes in fertility dynamics,
nonlinear outcome was used to decompose the effect of distinguishing it from other factors like socioeconomic or
increased contraceptive use on teenage pregnancy decline, demographic influences. It allows the breakdown of fertility
given that the outcome was a dichotomous variable. The rate changes into direct and indirect effects, providing a

Volume 11 Issue 5 (2025) 151 https://doi.org/10.36922/ijps.4866