Abstract / Summary
Ethiopia’s stagnating fertility transition has exposed the limitations of generalized public health strategies. This study aimed to move beyond national averages by identifying context-dependent determinants and evaluating the adjusted conditional associations of reproductive knowledge and media exposure on fertility preferences across distinct socioeconomic archetypes. Using cross-sectional data from the Small, Happy, and Prosperous Family in Ethiopia (SHaPE2) survey ( n = 16,322), we applied k-means clustering to stratify regional data into Rural, Transitional, and Urban archetypes. To adjust for high-dimensional confounders and capture non-linear heterogeneity, we utilized Double Machine Learning (DML), integrating LightGBM-based Linear DML and Causal Forest DML. Additionally, Robustness Values were calculated to quantify the threshold of omitted variable bias. The associations of these information interventions were context dependent. In the Rural cluster, Knowledge Depth ( \(\:-0.051;p<.001\) ) and Media Exposure ( \(\:-0.063;p=.009\) ) were correlated with reduced fertility preferences. Conversely, in the Transitional cluster, Knowledge Depth ( \(\:+0.067;p<.001\) ) and Media Exposure ( \(\:+0.101;p<.001\) ) were correlated with higher fertility preferences. In the Urban cluster, Knowledge Depth ( \(\:+0.048;p=0.002\) ) demonstrated a positive association, whereas Media Exposure ( \(\:-0.075;p=0.005\) ) correlated with lower fertility preferences. Causal Forest DML indicated that the association of media exposure depends on pre-existing knowledge in rural and transitional settings but operates structurally independently in urban environments. The assumption that information interventions are uniformly associated with reduced fertility demands has been flawed. In transitional clusters, increased reproductive knowledge and media exposure correlate with higher fertility preferences. Although the cross-sectional nature of our data restricts these findings to conditional associations rather than causal effects, they suggest that public health policies could benefit from shifting to cluster-specific precision strategies. This includes the persistence of foundational education in rural areas, prioritizing economic structural shifts over information campaigns in transitional zones and leveraging standalone mass media for normative reframing in urban centers.