Determinants of Government Health Expenditure in the European Union: Robust Panel Evidence from the EU-27, 2000–2023

Background: The government’s health expenditure in the European Union is determined by a number of factors, including economic capacity, financing arrangements, demographic pressures, health needs, environmental exposure, and labour-market conditions. The EU-27 provides a significant comparative context, given that Member States share a broad institutional framework while exhibiting considerable variation in fiscal capacity, population structure, and health-system financing. Methodology: This study analyses a balanced EU-27 panel for the years 2000–2023. The econometric approach addresses such phenomena as cross-sectional dependence, mixed integration orders, endogenous structural breaks, uncertain panel cointegration, and non-standard residual dependence. The primary stationary specification uses country and year fixed effects with Driscoll–Kraay standard errors. Supplementary robustness checks encompass alternative and predetermined financing measures, common-correlated-effects estimators, and finite-T dynamic sensitivity analysis. Results: Nine of the thirteen analysed series are I(1), while the remaining four are I(0). Bootstrap Westerlund tests do not provide evidence of robust multivariate panel cointegration. In the primary specification, GDP growth is positively associated with the annual log change in government health expenditure per capita, measured in current-PPP international dollars (β = 0.412, p < 0.001). Unemployment is negatively associated with expenditure change. Conversely, old-age dependency is positively associated with expenditure change, although this association is more sensitive to model specification. The strong contemporaneous coefficients on private health-expenditure share and government health budget share become negligible and statistically insignificant when one-period-lagged changes are employed. This finding suggests that these are contemporary financing-composition co-movements rather than prospective effects. Under pooled and heterogeneous common-correlated-effects estimators, the GDP coefficient remains positive but is substantially attenuated and loses conventional statistical significance. A finite-T sensitivity analysis reveals a weak positive short-run persistence; however, this is insufficient to substitute for the parsimonious static specification. Conclusions: Economic capacity is the most stable positive correlate across the financing-robustness exercises. However, common-factor adjustment reduces both its magnitude and statistical significance. The findings support cautious interpretation of the financing and demographic associations and demonstrate that estimator choice matters in EU health-expenditure panels.