BGBI: A citywide-referenced and bidirectional land use mix index for planning and policy evaluation

Source abstract

Land use mix (LUM) metrics provide actionable evidence to monitor planning goals and prioritize interventions, including transit-oriented development, zoning revisions, and urban retrofits. Yet common measures suffer from two main limitations that often render them difficult to translate into decisions. First, all measures summarize how mixed a place is without indicating the direction of the mix, i.e. whether a homogeneous area is predominantly residential or non-residential. Second, most metrics attain their maximum LUM value at a 50/50 split, even when the citywide composition is highly asymmetric. Here we introduce the Bidirectional Global-centered Balance Index (BGBI), a bounded signed index for land use composition that preserves directionality and centers interpretation on the observed citywide reference share. We compute BGBI alongside four widely used benchmarks and empirically validate all indices against job accessibility metrics by testing whether each measure’s implied balance target coincides with accessibility optima. The analysis uses point address data to measure LUM in terms of residential and non-residential shares aggregated in H3 hexagonal grids for three Brazilian cities. The full workflow can be executed using an open-source R package, which supports reproducible computation of all indices across multiple H3 resolutions and facilitates straightforward replication and sensitivity checks. Our results indicate that the highest-accessibility locations occur near baseline-centered balance (BGBI ≅ 0), rather than near the 50/50 reference implicit in many conventional metrics. These results show that interpretability properties can change diagnostic conclusions and suggest that baseline-centered directionality provides a practical target for planning and policy monitoring.