Public financing, external assistance and household health payments: a causal machine learning analysis of 173 countries, 2000–2023
Abstract
BackgroundOut-of-pocket (OOP) payments remain the dominant health financing mechanism in many countries while development assistance for health is contracting. Existing cross-country evidence on financing composition and household payments is weakened by an accounting identity linking regressors to the outcome and by the allocation of financing to high-burden settings.MethodsWe analyzed 4,141 country–year observations from 173 countries (2000–2023) using the WHO Global Health Expenditure Database and Sustainable Development Goal (SDG) indicators. Government and external shares of current health expenditure (CHE) were entered jointly, so each effect represents a reallocation from domestic private financing. The primary outcome was per capita real out-of-pocket spending. Double machine learning was applied across three specifications, culminating in a within-country design with three-year lags, pre-treatment covariates, and a baseline adjustment. Inference was country-cluster-robust.ResultsWithin countries, reallocating 10 percentage points from private to government financing was associated with 11.4% lower out-of-pocket spending per capita (95% confidence interval [CI: 7.9–14.8]), and reallocating to external financing was associated with 14.7% lower spending (11.2–18.0); the difference was not significant (p = 0.075). Conventional specifications overstated these effects by 2.6–2.9-fold. Share-based estimates matched the value implied by the accounting identity. Results were robust across all sensitivity analyses. No effects on tuberculosis, Human Immunodeficiency Virus (HIV), or maternal mortality were detected, and regional estimates did not support a policy typology.ConclusionShifting financing toward pooled government or external sources is associated with lower household payments, but by far less than cross-country estimates suggest. Financing targets and assessments of aid withdrawal should rely on evidence from within the country.
Related
- Why women’s health financing collapses when donors retreat and interest payments rise
- Out-of-pocket health expenditure and the health financing mix in Gulf Cooperation Council countries, 2000–2023: a descriptive analysis of WHO data
- The impact of male circumcision on risk of HIV infection in sub-Saharan Africa: a household fixed-effects analysis of 279,351 men from 29 countries.
- The Elephant in the Operating Room: Disclosure of Payments From Industry
- Accountability for pandemic financing decisions
BibTeX
@article{alraeesi2026public,
title = {{Public financing, external assistance and household health payments: a causal machine learning analysis of 173 countries, 2000–2023}},
author = {Roqaya Alraeesi and Rula Odeh Alsawalqa and Younas Khan},
journal = {Frontiers in Public Health},
year = {2026},
month = nov,
doi = {10.3389/fpubh.2026.1960014},
url = {https://www.frontiersin.org/articles/10.3389/fpubh.2026.1960014},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Public financing, external assistance and household health payments: a causal machine learning analysis of 173 countries, 2000–2023
AU - Roqaya Alraeesi
AU - Rula Odeh Alsawalqa
AU - Younas Khan
PY - 2026
DA - 2026/11/05
JO - Frontiers in Public Health
DO - 10.3389/fpubh.2026.1960014
UR - https://www.frontiersin.org/articles/10.3389/fpubh.2026.1960014
AB - BackgroundOut-of-pocket (OOP) payments remain the dominant health financing mechanism in many countries while development assistance for health is contracting. Existing cross-country evidence on financing composition and household payments is weakened by an accounting identity linking regressors to the outcome and by the allocation of financing to high-burden settings.MethodsWe analyzed 4,141 country–year observations from 173 countries (2000–2023) using the WHO Global Health Expenditure Database and Sustainable Development Goal (SDG) indicators. Government and external shares of current health expenditure (CHE) were entered jointly, so each effect represents a reallocation from domestic private financing. The primary outcome was per capita real out-of-pocket spending. Double machine learning was applied across three specifications, culminating in a within-country design with three-year lags, pre-treatment covariates, and a baseline adjustment. Inference was country-cluster-robust.ResultsWithin countries, reallocating 10 percentage points from private to government financing was associated with 11.4% lower out-of-pocket spending per capita (95% confidence interval [CI: 7.9–14.8]), and reallocating to external financing was associated with 14.7% lower spending (11.2–18.0); the difference was not significant (p = 0.075). Conventional specifications overstated these effects by 2.6–2.9-fold. Share-based estimates matched the value implied by the accounting identity. Results were robust across all sensitivity analyses. No effects on tuberculosis, Human Immunodeficiency Virus (HIV), or maternal mortality were detected, and regional estimates did not support a policy typology.ConclusionShifting financing toward pooled government or external sources is associated with lower household payments, but by far less than cross-country estimates suggest. Financing targets and assessments of aid withdrawal should rely on evidence from within the country.
SN - 2296-2565
ER -