Methodology

Where the numbers come from

Four official, public sources, each used for different things — led by the population and demographics that anchor every page.

Population, demographics, commute time & work-from-home — ACS 5-Year (2024)

Each city’s population, median age, median household income, home value and rent, education and poverty, plus its mean commute time (the basis for the Commute Score) and work-from-home share, come from the American Community Survey, 2024 5-Year Estimates (Tables B01003, B01002, B19013, B25077, B25064, B15003, B17001, B08303, B08301). The national homepage figures use the ACS 1-Year (Table S0801). We use the Census Bureau’s own published estimates rather than re-deriving them.

Race & ethnicity — ACS Table B03002

The race & ethnicity breakdown bars on each city page come from ACS Table B03002 (Hispanic or Latino origin by race). We use this table specifically because its categories are mutually exclusive and sum to the total population — White, Black, Hispanic or Latino (of any race), Asian, American Indian/Alaska Native, Native Hawaiian/Pacific Islander, two or more races, and other — so the shares add up cleanly and Hispanic origin isn’t double-counted against race.

The racial dot map — 2020 Census (P.L. 94-171)

The interactive racial dot map on each city page is built from the 2020 Census Redistricting Data (P.L. 94-171), Table P2 — the most detailed race & ethnicity count the Census Bureau publishes, available down to the individual census block. We place one dot for a set number of residents in each block (the exact figure is shown in the map legend) at that block’s geographic center, colored by race or ethnicity, using the same technique as the well-known national racial dot maps. Because it’s a full enumeration of every block rather than a survey, the map shows real neighborhood-level patterns; the Bureau applies privacy noise that can nudge very small blocks but leaves those patterns intact. Note the dot map (2020 count) and the breakdown bars (2024 ACS estimate) can differ slightly because they’re different vintages and methods — both are labeled on the page.

Commute & workforce flows — LEHD LODES (2023)

The flow maps, workforce counts, and home–work distances come from the Census Bureau’s LEHD Origin–Destination Employment Statistics (LODES), version 8, 2023 vintage (the latest available). For nearly every job covered by state unemployment insurance, LODES records the Census block where the worker lives and the block where they work, plus earnings and industry. We aggregate those block-level flows up to cities (Census Places), and we count cross-state commuters — people who live in one state and work in another (NJ→NYC, MD/VA→DC) — in the city they live in. The Census Bureau infuses statistical noise into LODES to protect privacy; aggregate patterns stay accurate. LODES documentation →

Internet availability — FCC National Broadband Map

The internet provider options on each city page come from the FCC National Broadband Map’s address-level serviceability data — the same source our partner FindBetterInternet uses, so the providers line up with theirs.

What “commute distance” means — and what it doesn’t

Every distance on CensusFlow is a straight-line (great-circle) distance between the geographic centroid of the home place and the centroid of the work place — “as the crow flies,” using the haversine formula.

It is not road distance and not travel time. LODES contains no travel time and no mode of transport — only where people live and where they work. A straight-line distance is always shorter than the real driving route, and it says nothing about how long a trip takes. We label it “straight-line home–work distance” throughout the site for exactly that reason. Read it as a consistent, comparable measure of how far apart home and job are — not as a door-to-door commute time.

What we leave out of the distance — and why

The headline distance is the median (and average) among residents who commute to a different place. Two groups are deliberately excluded from that calculation:

1. People who work where they live (~0 miles)

Many residents work in the same city they live in, so their home-to-work distance is essentially zero — and in large cities they can be the majority. Leaving them in would collapse the “commute distance” toward 0 and misrepresent the trip for people who actually travel. So we exclude same-place workers from the distance metric; they’re already represented separately as “% who work locally.”

2. “Commutes” longer than 60 miles (data artifacts)

LODES sometimes assigns a worker to their employer’s headquarters or mailing address rather than the building they actually report to. That creates a handful of impossible 100–500-mile “commutes” — a worker shown traveling clear across the state — that aren’t real daily trips. Because those flows carry real worker counts, they can dominate a median once same-place workers are removed. We therefore exclude any flow beyond 60 straight-line miles from the distance metric. Sixty miles is generous for a genuine long commute while reliably filtering out these employer-address artifacts. This exclusion affects only the distance figures — not the maps, the destination/origin lists, or any other statistic.

Median vs. average

We lead with the median — the middle value, where half of commuters travel less and half travel more. The median resists distortion from a few unusually long trips, so it’s the most honest single number for “a typical commute here.”

The average (mean) is the total distance divided by the number of commuters. It’s useful but more sensitive to outliers, which is why we don’t headline it. Both are computed the same way: over residents who commute to a different place, within the 60-mile ceiling above.

The Commute Score

Each city page leads with a Commute Score from 0 to 100 that answers one question: how hard is the daily commute for someone living in or around this city? Higher is better — a high score means short commutes; a low score means long, congested ones.

It’s based on the city’s mean travel time to work from the ACS (2024 5-Year, Table B08303) — real reported commute minutes, not a straight-line distance — mapped onto an absolute 0–100 scale:

Scores between those points are interpolated. The ring is colored by tier: 80–100 strong green, 60–80 light green, 40–60 yellow, below 40 red.

Because the scale is absolute, not a ranking, the score reflects a city’s actual commute burden rather than its size. A congested metro like New York (about a 40-minute mean commute) lands in the low 20s — red — while a quick-commute town near 15 minutes lands in the high 80s — green. Cities below the ACS reporting threshold for commute time are left unscored rather than shown a misleading number.

Internet options

The internet providers on each city page come from the FCC National Broadband Map, using the same method as our partner FindBetterInternet so the numbers line up. For each city we take broadband-serviceable availability (fiber, cable, and DSL at ≥25/3 Mbps; satellite and fixed wireless excluded), roll it from Census tracts up to the city’s ZIP codes, and weight each area by its residential share (the HUD ZIP–tract RES_RATIO) so coverage reflects where people actually live. The top providers are ranked by that residence-weighted coverage; we show the names and the fastest plan available, not a coverage percentage.

We also show a speed-availability breakdown — the residence-weighted share of a city’s homes that can buy a wired plan at 1, 2, 5, and 8 Gbps (advertised download speeds). Because gigabit service is now near-universal in the places we cover, the 1 Gbps row is a muted baseline; the higher tiers are highlighted, because that’s where availability genuinely differs from one city to the next.

The “Fiber” badge marks genuine fiber service only. It comes from a curated list of fiber providers — and, for mixed copper/fiber telcos, only where fiber is actually present in that city — never from cable providers (the raw data over-labels some cable footprints as fiber). A full plan comparison is one click away via the FindBetterInternet link.

“Total Workforce” vs. “Workers Living Here”

These two numbers measure different things, so they won’t match — and that’s expected:

• Total Workforce (on a state page) counts jobs located in that state — every job physically in, say, Texas, no matter where the worker lives. A workplace count.
• Workers Living Here (on a city page) counts employed residents of that city — people who live there and have a job, no matter where that job is. A residence count.

A state’s job total and the sum of its cities’ resident-worker counts describe two different universes (jobs vs. residents, whole state vs. covered cities), so they aren’t meant to add up to each other.

Which places CensusFlow covers

CensusFlow publishes pages for about 1,160 U.S. cities — not every incorporated place, and no ZIP-code pages. The selection combines two rules: every state gets at least its top 10 cities by employed residents (so even smaller states are browsable), plus any city that ranks in the national top 1,000 by size (so large states contribute many more). “Cities covered” on a state page means the cities we publish from that state, not the total number of places in it. Focusing on the largest places keeps every page resting on a solid sample of LODES data.

Earnings tiers

LODES splits workers into three monthly-earnings bands, which drive the map’s earnings colors:

From Census blocks to cities

LODES is recorded at the Census block level. We roll those blocks up to Census Places (incorporated cities, towns, and CDPs) using the LODES crosswalk, and every CensusFlow page is a Census Place. We don’t publish ZIP-code pages — ZIP Code Tabulation Areas only approximate USPS ZIPs, and cities are the clearer, more accurate unit for commute patterns.

Crucially, a Census Place is the city proper — the incorporated city within its legal limits, not the surrounding metro area. So every figure on a city page (workers, commute distance, industries) covers just that city, and will read smaller than the metro-area numbers many sites quote — a metro bundles in dozens of neighboring cities, suburbs, and counties. We report the city itself, on purpose.

The interactive maps

Each map is a dot-density view where one dot ≈ 75 workers, scattered within the home or work area to show density without implying exact addresses. You can switch between “Where residents work” and “Where workers live,” and color the dots by earnings tier or industry. Maps render with MapLibre GL JS and deck.gl. The dot pattern illustrates volume and direction — it is not a precise plot of individuals.

Known limitations

• Which jobs are counted. We use LODES “All Jobs” (JT00), which covers private and government work — state, local, and federal-civilian employment are all included. Two groups fall outside LODES, which is built from payroll/unemployment-insurance records: active-duty military and the self-employed / independent contractors. So heavily military towns understate their uniformed population, and areas with lots of self-employment are undercounted.
• Straight-line, not road distance. The home–work distances are straight-line and understate real driving distance. For the actual commute burden we use ACS reported travel time (the Commute Score), which captures congestion that distance can’t.
• Employer-address artifacts. Some workers are tied to an employer’s HQ/mailing address rather than their actual worksite. Excluding flows over 60 miles removes the worst of these, but shorter mis-assignments can remain.
• Older LODES data for some states. A few states publish LODES on a lag, so their latest vintage is earlier than 2023 — notably Alaska (2016) and Michigan (2021). Those pages are labeled with their data year, and we show fewer of their cities (their largest only). The Commute Score is unaffected — it uses current ACS commute time.
• Privacy noise. LODES is privacy-protected with injected statistical noise — accurate in aggregate, but not for any individual worker or employer.

Source data: U.S. Census Bureau, LEHD LODES (2023) and American Community Survey (2024 5-Year); FCC National Broadband Map. CensusFlow is an independent project and is not affiliated with the U.S. government.