Technical overview
GONI is computed using a nested, hierarchical spatial comparison framework based on a global hexagonal grid.
All calculations are performed on spatial units that are independent of administrative boundaries and are, conditional on location, as-good-as-randomly assigned with respect to socio-economic characteristics.
This grid-based construction helps reduce bias introduced by endogenous political or statistical boundaries. Because spatial units are defined independently of administrative, political, or statistical boundaries, comparisons are not driven by boundary choice.
Conditional on geographic location, neighbourhood assignment is as good as random, supporting interpretation of differences as reflecting underlying spatial structure rather than artefacts of zoning or jurisdiction.
Step 1: Metric construction at neighbourhood level
For each neighbourhood cell, a set of raw indicators is measured (e.g. population, economic activity, service density).
Because neighbourhood boundaries are defined by a fixed global grid rather than policy or census design, assignment of observations to neighbourhoods is exogenous to local socio-economic outcomes, up to geographic location.
Step 2: Aggregation to higher spatial levels
Neighbourhood-level metrics are aggregated upward using population-weighted means:
- neighbourhood → local (r8 → r6)
- local → regional (r6 → r4)
This aggregation preserves population concentration while maintaining spatial comparability across contexts.
Step 3: Nested percentile computation
Often, spatial indicators are presented in absolute terms that are difficult to interpret in isolation
—for example, is 700 people per km² dense or sparse? Accordingly, percentile ranks are computed separately at each spatial level, using scale-appropriate reference sets:
Neighbourhood percentile (r8-in-r6):
Neighbourhoods are ranked relative to other neighbourhoods within the same local area.Local percentile (r6-in-r4):
Local areas are ranked relative to other local areas within the same region.Global percentile (r4-in-world):
Regions are ranked relative to all regions worldwide.
Each percentile therefore reflects position within a comparison group that is not mechanically determined by administrative design.
This is particularly advantageous in remote areas, where comparison sets are naturally more restricted, yet administrative boundaries may place them into distant or heterogeneous groupings for administrative convenience.
Step 4: Propagation of contextual scores
Regional and global percentiles are mapped back down to neighbourhood cells.
Each neighbourhood therefore carries:
- its own local performance score,
- the strength of its surrounding local environment, and
- the global position of the region it belongs to.
Step 5: Composite index construction
The Global–Ordinal Neighbourhood Index (GONI) is constructed as a weighted combination of the nested percentile scores.
All components lie in the interval ([0,1]), weights are normalised, and the resulting index preserves ordinal interpretation at each scale.
Index percentiles are recomputed using the same nesting logic to ensure internal consistency.