Demographic coverage Planner

Population, households, workers, jobs, and equity demographics within walking distance of your stops, tallied at the individual census-block level from US Census data. Per-route and system-wide figures suitable for a grant narrative ("33,600 residents and 32,000 jobs are within a quarter-mile of the network") or for sanity-checking a service alternative.

The system-level summary is free on every plan. The per-route coverage breakdown is a Planner feature.

What it is

The Coverage panel reports, for the area within walking distance of your transit network: population, households, workers, and jobs, plus two demand estimates, ridership propensity and transit need, built from four underlying segments: people in zero-vehicle households, people under 200% of the federal poverty line, seniors 65+, and people with a disability. Alongside the totals it shows an equity profile (minority, low-income, zero-vehicle, senior, and youth shares) compared against the county baseline, and a Download CSV of every figure.

The counts are exact rather than approximate: GTFS·X tallies the individual US census blocks whose center falls inside the walkshed. Block population comes from the 2020 decennial count, American Community Survey demographics are apportioned down to each block by population, and jobs are workplace counts from the Census LEHD program. It works for any feed in the 50 states plus DC.

It answers the most-asked question in a service-planning meeting ("how many people does this route reach?") without firing up a desktop GIS or pulling Census tables by hand. Run it on your existing network to anchor a discussion, or on a proposed alternative to support a recommendation.

When to use it

• Grant applications. Most FTA discretionary programs (5310, 5311, 5339, Areas of Persistent Poverty, etc.) ask for population served. A defensible number with a documented methodology beats a hand-waved estimate every time.
• Service-planning workshops. Comparing two corridor options? Run coverage on both and put the numbers side by side. The conversation shifts from "I think A is better" to "A reaches 4,000 more households but 1,200 fewer jobs."
• NTD reporting context. Population in service area is one of the NTD reportable metrics; the Coverage figure is a defensible input.
• Title VI program updates. A baseline-coverage figure pairs naturally with the Title VI analysis — the latter is a ratio, the former is the denominator.

How to use it in GTFS·X

1. Open the feed in the editor and click Coverage in the left sidebar.
2. Click Analyze Coverage. GTFS·X loads the census blocks around your stops and tallies the ones inside the walkshed. The first run for a region takes a few seconds; results cache locally.
3. Read the system summary: population, households, workers, jobs, and the ridership propensity and transit need estimates, with the count of census blocks covered noted underneath.
4. Below it, the demographic profile compares the covered area's equity shares against the county baseline (a ratio above 1.0 means the served area over-represents that group).
5. Scroll the per-route breakdown to see how each route contributes, then use Download CSV to export every figure (system, county baseline, and per route) for a report. Pair a route's coverage with the Cost estimation panel for a service-per-dollar view.
6. Planner-plan users can switch on street-network walksheds for walk-distance reach instead of a straight-line buffer.

Methodology

GTFS·X counts the individual US census blocks whose center point falls inside the walkshed, then sums each block's attributes. Counting real blocks rather than approximating with overlap math is what makes the figures exact.

• Walkshed. By default, a straight-line buffer around each stop: quarter-mile (1,320 ft) for bus, ferry, cable car, gondola, and funicular routes; half-mile (2,640 ft) for tram, subway, and rail. These match the conventional walking-distance assumptions used in transit-planning practice. Planner-plan users can switch to street-network walksheds, which follow the actual pedestrian network (and can size each stop's reach by its service frequency) instead of a circle.
• Census blocks, not block groups. A census block is the Bureau's smallest geography, often a single city block. GTFS·X loads a nationwide layer of every block's internal point and tallies the blocks inside the walkshed, so coverage tracks your stop placement closely instead of being smeared across a half-mile disc. (An earlier version of this tool approximated block groups as discs on tract centroids; the block-level method replaces it.)
• Data sources and vintages. Block geometry comes from the Census TIGER/Line 2025 block file (TABBLOCK20). Demographic detail comes from the American Community Survey 5-year estimates, published at the block-group level and apportioned down to each block by a scalar weight per block: 2020 decennial POP20, falling back to HOUSING20, then ALAND20, then an even split, wherever population data is unusable. Jobs come from the Census LEHD/LODES program, which already reports employment at the block level. The ACS vintage is probed rather than hardcoded, so this layer, the Demand Dots tiles, and this page all run against the same release; as of this writing that's ACS 2020–2024 5-year and LODES 8 (2023). Connecticut is pinned one ACS year back (2021), because TIGER still codes Connecticut's blocks to the counties the state retired in 2022.
• Population, households, workers, jobs. Population (ACS B01003), households (occupied households, ACS B25044), and workers (employed residents counted where they live, ACS B08301) are residence-based. Jobs are workplace-based, from LODES, and are never added to a residence-based figure. The equity profile reports the minority, low-income, zero-vehicle, senior, and youth shares of the covered area against the surrounding county.
• The four segments. Carless: people in zero-vehicle households (ACS B25044), converted from households to people using the PUMA's own measured average size for zero-vehicle households specifically, about 1.8 people, rather than the roughly 2.43 average for all households; zero-vehicle households skew smaller (more single-person and elderly households), so the overall average would overcount. Low income: people under 200% of the federal poverty line (ACS C17002). Seniors: adults 65+ (ACS B01001). Disability: civilian adults 18+ living with a disability (ACS C21007); the tables that would extend this to children or people in institutions (B18101, B18105, C18108) are tract-only and not usable at block group. Every segment is a straight ACS count, apportioned to the block the same way population is; the panel badges each one ACS count.
• Ridership propensity and transit need. Ridership propensity is the union of carless and low income. Transit need is the union of all four segments. Both are deduplicated: a person who is carless and low income counts once, not twice. Published ACS tables give only the count in each segment, never the overlap between them, so the union can't be read off the tables directly. GTFS·X starts from an independence backbone (the block's population times one minus the chance, assuming independence, that a person is in none of the segments), applies a correction factor measured from Census PUMS (2020–2024 5-year) person microdata for that block's PUMA, and clamps the result so it can never fall below the largest single segment or exceed the sum of the segments (or the block's population). Held out against PUMS, this runs about 1.66% MAPE for propensity and about 2% for need, versus 34% for the retired flat-0.6 model. The panel badges both estimate; neither should be added to the segments or to each other, since need already contains everything in propensity. Full derivation and citations: Rider propensity.
• System vs. per-route. The system summary counts each block once even when several routes reach it (a union, not a sum). Per-route figures are pure per-route, so adding them up will over-count blocks served by more than one route.
• Coverage area. The exact block layer spans the 50 states plus DC. A feed outside that (a US territory, or outside the country) automatically falls back to a block-group estimate read live from the Census API at each block group's tract centroid, and the panel labels which method produced each figure. That fallback reports population, households, workers, jobs, and the equity profile as plain ACS counts; it does not report ridership propensity or transit need. A tract-centroid apportionment is the wrong instrument for a deduplicated union estimate, so the panel simply doesn't show one outside the exact block layer.

Limits

• The default straight-line buffer treats stop placement as the unit of access; it doesn't model walk-network geometry, intersections, or topography. A stop on the wrong side of a freeway with no underpass still counts as covering everything within the buffer. The Planner-plan street-network walkshed addresses this by following the real pedestrian network.
• ACS figures are published at block group, not block, so every one of them—population included—is apportioned down to blocks by a per-block weight (2020 decennial block population, falling back to housing units, then land area). That is far more accurate than a land-area split, but it is still an approximation at the block level: it assumes the block group's people are distributed like its 2020 population was. The one figure measured natively at block level, and so not apportioned at all, is jobs (LODES is published per block). Ridership propensity and transit need carry one further layer of approximation on top of the apportionment: the PUMA-level correction described above.
• Ridership propensity and transit need are the same PUMS-corrected model used by the Rider propensity map layer, and inherit its limits: there's no published joint distribution below the PUMA level, so the deduplicated union can't be checked against ground truth at the block level; race and ethnicity are deliberately excluded as a predictor (use the Title VI analysis panel for equity work); and small-area ACS estimates carry real sampling error, most visible in sparsely populated blocks. See Rider propensity for the full derivation, citations, and validation numbers.
• "Workers" and "jobs" measure different things and are both reported: workers are employed residents counted where they live; jobs are counted where the work is (so a downtown with few residents but many jobs shows a low worker count and a high job count). Jobs are a workplace universe and are never combined with a residence-based figure.
• Figures use the current auto-probed ACS 5-year vintage (2020–2024 as of this writing) and TIGER 2025 block geometry; both refresh automatically each year, typically each January after the new ACS release. If your agency's adopted methodology requires a specific vintage, cite the release noted on this page as of your report date.
• The on-screen equity profile is a quick coverage-vs-county comparison. For a formal Title VI service-equity analysis (disparate-impact and disproportionate-burden tests), see the Title VI analysis page.