3.2 Spatial regression estimates
Results of spatial regressions show a good explanatory power in terms of adjusted
R2, particularly for primary schools, with failed rejection of the hypotheses of residual
non-normality and serial correlation (Table 2; residual autocorrelation might
arise due to the cross-district sample design, given geographically contiguous observations).
Neither the district population (results not shown), nor its location
in terms of agro-climatic zone appear to substantially matter for changes in access
to public healthcare, while they do for the respective changes in access to primary
education. Unlike results of previous (Tobit: see note 1) model estimates, household
size does not turn out to have a significant explanatory role for healthcare.