Methodology

Poverty is measured using the Foster–Greer–Thorbecke class: headcount ratio (FGT0), poverty gap (FGT1) and squared poverty gap (FGT2), computed against a poverty line you specify in the survey's own welfare units. Inequality is reported as the Gini coefficient. Where a survey weight variable is supplied it is applied throughout, so estimates are population-representative rather than sample averages.

Spatial autocorrelation uses Moran's I with row-standardised Queen contiguity weights, computed via PySAL. Local clustering uses Local Moran's I (LISA), with quadrants reported as high-high, low-low, high-low and low-high; a unit is labelled a cluster only where the local statistic is significant at the 5% level. Administrative boundaries are sourced from geoBoundaries under CC-BY 4.0 where you do not supply your own.

Moran's I and LISA require at least five matched geographic units; below that the platform reports the statistic as unavailable rather than computing something unstable.

Country indicator series are harmonized from their providers and presented with the years each series actually covers, so gaps are visible rather than interpolated away. AIC does not impute missing observations or extend series beyond their source.

Estimates are not smoothed, back-cast or reconciled between sources. Where two providers disagree, the provenance of each figure is shown rather than a blended value presented as fact. Coverage varies by country, indicator and year according to what the source publishes.