Areal Disaggregation: A Small Area Estimation Perspective

Abstract

Producing reliable estimates of health and demographic indicators at fine areal scales is crucial for examining heterogeneity and supporting localized health policy. However, many surveys release outcomes only at coarser administrative levels, thereby limiting their relevance for decision-making. We propose a fully Bayesian, single-stage spatial modeling framework for area-level disaggregation that generates fine-scale estimates of indicators directly from coarsely aggregated survey data. By defining a latent spatial process at the target resolution and linking it to observed outcomes through an aggregation step, the framework adopts small-area estimation techniques while incorporating covariates and delivering coherent uncertainty quantification. The proposed methods are implemented with inlabru to achieve computational efficiency. We evaluate performance through a simulation study of general fertility rates in Kenya to demonstrate the models' ability to recover fine-scale variation across diverse data-generating scenarios. We further apply the framework to two national surveys to produce district-level fertility estimates from the 2022 Kenya Demographic and Health Survey and, more importantly, district-level indicators for unpaid care and domestic work and mass media usage from the 2021 Kenya Time Use Survey.

Figures (11)

• Figure 1: Comparison of disaggregation and benchmark models (FH and direct estimates at the Admin-2 level) across four simulation scenarios and evaluated with within-Admin-1 regression, $R^2$; Pearson correlation, $r$; the uncertainty IS; width; and coverage. Scenario 1 includes observed covariates and random effects at the individual and area levels. Scenario 2 expands scenario 1 with the introduction of two additional area-level measures that are assumed to be unobserved. Scenario 3 expands scenario 2 with the addition of IID Admin-2 level random effects. Scenario 4 is the most complicated scenario and introduces model misspecification by allowing the coefficients to differ between urban and rural areas within the same Admin-2 unit.
• Figure 2: Admin-2 level estimates of GFR in Kenya based on DHS 2022 data and obtained from four models: direct estimation, FH at Admin-2, FH disaggregation, and unit-level disaggregation with MRP. The upper panel shows point estimates of GFR, and the bottom panel shows the width of the 90% uncertainty intervals.
• Figure 3: Estimated proportion of the day that women spend on unpaid care and domestic work from direct estimation at Admin-1, FH disaggregation, and unit-level disaggregation models. The maps show both point estimates and the widths of the associated 90% uncertainty intervals.
• Figure 4: Maps of the female-male gap in unpaid care and domestic work and the associated probability that the difference exceeds four hours per day. Admin-1 estimates are based on direct estimates, and Admin-2 estimates are from a unit-level disaggregation model with MRP.
• Figure 5: Maps of the female- and male-specific estimates for mass media usage.