Kauzální vliv zdravotnických zdrojů na úmrtnost: mezinárodní komparativní analýza

This thesis examines the effect of healthcare resources on mortality in a panel of 127 low-, lower-middle-, and upper-middle-income countries observed from 2000 to 2018. The analysis focuses on physician density and per capita health expenditure as the main healthcare inputs, while infant mortality, under-five mortality, and maternal mortality are used as the outcome variables. To address non-stationarity and unobserved country heterogeneity, the empirical strategy combines first-differenced two-way fixed-effects models with Driscoll--Kraay standard errors and a Double/Debiased Machine Learning estimator for static panel data, implemented with four flexible learners. The results provide the strongest evidence for a negative relationship between per capita health expenditure and under-five mortality. In the DML estimates, the LASSO coefficient on log per-capita health expenditure for under-five mortality is approximately -0.04 and statistically significant at the five percent level, while the elastic net and random forest estimates are similar in sign and magnitude but only marginally significant. For infant mortality, the estimated health-expenditure coefficients are also negative across learners, but they do not reach conventional five-percent significance levels. Physician density, by contrast, does not provide robust evidence of mortality-reducing effects once flexible functional forms are allowed. Moreover, the positive physician-density coefficient is inconsistent with theoretical expectations and may reflect remaining endogeneity, including reverse causality. The interaction models likewise do not support the hypothesis that physician density becomes more effective when supported by greater nursing availability or higher health expenditure. Overall, the findings suggest that health financing is more consistently related to short-run improvements in child mortality than physician availability.