A four-part research sequence on grocery proximity, transit-stop coverage, housing tenure, and race in California.
This hub is the transit-equity research sequence behind Too Early To Say. It shows how an applied economist can examine grocery proximity alongside transit-stop coverage, housing tenure, and race. Each article is one part of the same California tract-level measurement problem. The agent handles routine source checks and data-pipeline tasks. The access definition, comparison, and interpretation remain ours.
This sequence is for transportation researchers, public agencies, and policy analysts who need to distinguish a documented input from an article-reported result. The problem it solves is simple. A tract classification can look precise while an incomplete feed inventory or unsupported threshold changes the population being described.
One idea runs through all four analyses. A transit-access claim is only as credible as the stop inventory and threshold that define it. A documented feed crosswalk supports an audit. An unresolved input leaves the resulting classification under reconciliation.
The sequence moves from descriptive comparisons to the transit input that supports them. In practice, reconciliation can send any analysis back to its source inventory.
Outcome: a tract-level housing-tenure comparison with its evidence status attached.
In the housing-tenure case, the article reports a 0.58-mile mean grocery distance for renter-dominated tracts. The value remains under reconciliation. The agent can audit the tract joins and source fields. We decide whether the comparison supports a substantive interpretation.
Outcome: a race-and-access comparison with conflicting results identified.
In the race-and-transit case, the article reports a 0.72-mile mean grocery distance for majority-minority tracts. The tract inventory remains under reconciliation. The agent can compare populations and table outputs. We decide which claims survive the conflict across pages.
Outcome: a raw-to-unique stop inventory with validation caveats.
In the Cal-ITP data-quality case, the pinned artifact labels 64,060 stop records as raw. That conflicts with an earlier article label. The agent can compare field definitions and deduplication steps. We decide whether the input population is documented well enough to classify tracts.
Outcome: a stop-proximity classification with project-defined thresholds.
In the mobility-desert case, the project applies its 0.5-mile rule to transit-stop proximity around grocery stores. The agent can apply the recorded spatial rule. We decide whether that rule answers the policy question.
The housing-tenure and race pages ground the descriptive comparisons. The data-quality page documents the unresolved raw-to-unique stop record. The mobility-desert page defines the project’s stop-proximity screen. None currently links a matched public output that reproduces its statewide result, so the pages identify article-reported claims and reconciliation requirements rather than a verified statewide estimate.
Primary inputs include the Cal-ITP GTFS-Ingest Pipeline Dataset and the USDA Food Access Research Atlas documentation. The public claim manifest records which project claims remain under reconciliation.