This analysis substantially revises and extends this modelling approach in three
directions, with a view to testing whether (i) revealed eligibility thresholds overor
understate the average rates of access coverage, thus not closely reflecting BIA
assumptions, (ii) the geographical diffusion of service provision is in line with reducing
the number of uncompensated losers as an additional social welfare objective
(Bentham’s numbers effect, which in the income distribution literature is equated
to numbers below the poverty line, and here to numbers remaining without access
to public services), and (iii) marginal improvements in public service delivery are
characterised by spatial heterogeneity, thus implying biased and inefficient OLS regressions.
With a reapplication to the above BIA case study, the analysis therefore
queries an a priori use of average accessibility as a universal yardstick (to distinguish
worse off from better off target areas), an exclusive focus on distribution and efficiency
as social welfare objectives, and unique functional relationships across regions
of a country (which imply uniform spatial dependence without local variations). The
remainder of this paper proceeds as follows: in section 2, theoretical elements are
reassessed and linked to the econometric modelling with global spatial regressions
and geographically weighted regressions (GWRs); section 3 reports and discusses
the econometric results for primary education and healthcare in Niger, while section
4 draws concluding remarks.