Map Explorer
Transit Gap — Suppressed Demand
The gap between predicted and actual transit share. Warm-colored tracts (orange-red) are where the Random Forest model expects significantly higher ridership than observed — areas where access constraints, not lack of demand, are suppressing ridership. These are the highest-priority zones for targeted investment. Cool or neutral tracts are either well-served or accurately predicted by the model.
Actual Transit Share
Observed transit commute share from ACS 2024 — the share of employed residents who travel to work by public transportation. Concentrations in Hudson, Essex, and Union counties reflect dense bus and rail service. Suburban and exurban tracts in Bergen, Morris, Somerset, and Middlesex show sharply lower shares, where car dependence dominates.
Predicted Transit Share
Random Forest model estimates of transit share for all 2,181 tracts, based on infrastructure access (dist_to_bus, dist_to_rail, bus_density_2mi) and demographics. Compare this map to Actual Ridership: where predicted values exceed actual values, the model is identifying suppressed demand — areas where infrastructure is the binding constraint, not lack of need. These tracts form the investment priority map.
Bus Stop Density — 2-Mile Buffer
Count of NJ Transit bus stops within 2 miles of each tract centroid. This was the single most important predictor in both OLS and Random Forest models, accounting for approximately 58% of Random Forest feature importance. The map reveals the dramatic contrast between urban cores (hundreds of stops within reach) and outer suburbs (fewer than ten). This density gap directly maps to the ridership gap.