Methodology

Principles

• Only official publishers, and only datasets whose licence allows reuse. The full list is on the sources page.
• No crowdsourced, scraped or modelled figures, and no gap is filled with a guess.
• Each figure keeps its period, its unit and its source. Where the publisher gives a margin of error, it is shown.
• Where a publisher gives no figure, the site says so and gives the reason: the figure is withheld, the sample is too small, or the dataset stops at a larger area.

Areas

The United States is shown at four levels: the country, the 50 states with the District of Columbia and Puerto Rico, 3,222 counties and county equivalents, and about 85,000 census tracts. Boundaries are the U.S. Census Bureau’s cartographic boundary files, which are simplified for mapping and are statistical areas, not legal property lines. The island areas (American Samoa, Guam, the Northern Mariana Islands and the U.S. Virgin Islands) are not included because the main sources do not cover them.

Land area comes from the boundary files. Areas are given in square kilometres, with square miles alongside.

What is done to the figures

Survey estimates and margins of error

Figures from the American Community Survey are estimates from a sample pooled over five years, so they describe a period (2020 to 2024), not a single year. Each comes with a margin of error at the 90 percent confidence level, printed after the figure with a plus-minus sign. For census tracts the margin is often large; read small differences between tracts with caution.

Open-ended and missing values

The Census Bureau reports very high or very low medians only as “at or above” or “at or below” a limit. These are shown with a plus sign or the words “or less”, never as an exact amount. Codes that mean “no estimate” are not turned into numbers: the figure is shown as not available.

Withheld cells

The Bureau of Labor Statistics withholds wage and employment cells that could reveal a single employer. They appear as “withheld”. They are not estimated from neighbouring cells or totals.

Calculated figures

A few figures are simple calculations from published ones, and are labelled as such wherever they appear:

• Population density. Calculated by Marsadata from the ACS 5-year population estimate (table B01003) divided by the land area in the Census Bureau's 2025 cartographic boundary files.
• Average home value by type of home. Calculated by Marsadata from ACS tables B25080 (total value of owner-occupied homes by units in structure) and B25032 (owner-occupied homes by units in structure): total value divided by the number of homes.
• People speaking a language other than English at home. Calculated by Marsadata from ACS table C16001: population aged 5 and over minus people who speak only English at home, divided by the population aged 5 and over, times 100.
• Business establishments per 1,000 residents. Calculated by Marsadata from County Business Patterns 2023 total establishments divided by the ACS 5-year 2020–2024 population estimate, times 1,000.

Money is shown in the source currency and is not adjusted for inflation.

Pay per hour, week or month

The Bureau of Labor Statistics publishes pay as an average weekly wage, and that is the figure the site stores. You can read it per month or per hour instead. Both are plain conversions made by Marsadata and are labelled as such: the monthly figure is the weekly wage times 52, divided by 12; the hourly figure is the weekly wage divided by 40 hours. The source does not report hours worked and counts part-time jobs like full-time ones, so the hourly figure understates actual hourly pay where part-time work is common. The classes on the map and the position of a place on a strip do not depend on the unit.

Home size and type of home

No official source for the United States publishes the floor area of homes together with their rent or value, so the site shows no rent or price per square metre for U.S. places. It will show them for countries whose statistics publish them. For the United States it shows the official breakdowns that exist:

• Median rent by number of bedrooms, published by the Census Bureau, so that a studio is compared with a studio and a three-bedroom home with a three-bedroom home.
• Home value for houses and for apartments. The Census Bureau publishes the total value of owner-occupied homes by type of building and the number of such homes; Marsadata divides one by the other. The result is an average, not a median, so a few very expensive homes pull it up and it is higher than the median value shown above it. The Bureau publishes no median by type of home.

Rent by type of building is not shown: the published totals cover only renters who pay cash rent, while the published counts of homes by type of building cover all renters, so dividing one by the other would give a figure that is too low.

The map

Each map layer groups areas into seven classes holding the same number of areas (quantiles), computed separately for states, counties and tracts on the latest period. The class boundaries are printed under the legend. The same seven classes are used for the strips on place and comparison pages, so the marker of a place shows which seventh of comparable areas it falls in. Areas with no published figure are drawn in grey.

The level drawn depends on the zoom: states, then counties, then census tracts. A figure that is published only for larger areas stays at that level when you zoom in.

What each figure is, and is not

• Population (people). Total resident population: American Community Survey 5-year estimate for 2020–2024 (table B01003). Five-year pooled sample survey (2020–2024), not a single-year figure; small areas carry large margins of error. It is a survey estimate, not a census count. The national figure covers the 50 states and the District of Columbia; Puerto Rico is published separately.
• Population density (people per km²). Residents per square kilometre of land area. Land area excludes water. The national figure uses the land area of the 50 states and the District of Columbia, matching the population estimate.
• Median home value (USD). Median value of owner-occupied homes: American Community Survey 5-year estimate for 2020–2024 (table B25077). The value is each owner's own estimate of what the home would sell for, not a sale price. Five-year pooled sample survey (2020–2024), not a single-year figure; small areas carry large margins of error. Medians above $2,000,000 are published only as "$2,000,000+" and medians below $10,000 as "$10,000-".
• Average home value by type of home (USD). Average value of owner-occupied homes, for all homes and by type of home: houses, apartments and mobile homes. It is an average, not a median: the total value of the homes divided by their number. American Community Survey 5-year estimates for 2020–2024 (tables B25080 and B25032). This is an average, not a median, so a few very expensive homes pull it above the median home value. Values are each owner's own estimate of what the home would sell for, not sale prices. "Apartment" means an owner-occupied home in a building with two or more units, condominiums and co-ops included; rented homes are not counted. The Census Bureau publishes no median value for houses or for apartments (only for mobile homes), which is why this figure is an average. It publishes the underlying totals for the country, most states and about a quarter of counties; no figure is shown where an input is missing. It is not shown for census tracts, where the totals are rare and too unstable to read without a margin of error. Five-year pooled sample survey (2020–2024), not a single-year figure; small areas carry large margins of error.
• Median gross rent (USD per month). Median monthly gross rent of renter households paying cash rent, for all rented homes (table B25064) and by number of bedrooms (table B25031): American Community Survey 5-year estimates for 2020–2024. Gross rent is the contract rent plus utilities. It covers all renter households, sitting tenants included, so it lags the rents asked for new leases. Five-year pooled sample survey (2020–2024), not a single-year figure; small areas carry large margins of error. Medians above $3,500 are published only as "$3,500+" and medians below $100 as "$100-". Medians by number of bedrooms rest on fewer households, so they are missing more often and less precise.
• Language spoken at home (people aged 5 and over). People aged 5 and over by the language they speak at home: English only and twelve language groups. American Community Survey 5-year estimate for 2020–2024 (table C16001). Sample survey: every cell has a margin of error, large for census tracts. Languages are published as twelve groups plus English only; individual languages are not available for counties and tracts.
• People speaking a language other than English at home (% of people aged 5 and over). Share of people aged 5 and over who speak a language other than English at home. Says nothing about English ability: many of these people also speak English very well. Based on a sample survey with large margins of error for census tracts; no margin of error is given for the share.
• Average wage (USD per week). Average wage per job, by place of work. The source publishes it as a weekly figure, the annual average of weekly wages: Quarterly Census of Employment and Wages. The total covers all industries and all employers, private and government; the sector breakdown covers the private sector only. Covers jobs at employers subject to unemployment insurance, more than 95% of U.S. jobs; the self-employed are not included. Wages include bonuses. It is an average per job, full-time and part-time alike, not per worker. Figures are withheld where they could reveal an individual employer. Recent figures are revised by the publisher.
• Employment (jobs). Annual average number of jobs, by place of work: Quarterly Census of Employment and Wages. The total covers all industries and all employers, private and government; the sector breakdown covers the private sector only. Counts jobs at employers subject to unemployment insurance, more than 95% of U.S. jobs; the self-employed are not included. A person with two jobs is counted twice. Figures are withheld where they could reveal an individual employer.
• Business establishments (establishments). Number of business establishments with paid employees, in total and by NAICS sector: County Business Patterns 2023. An establishment is a single physical location, so a company with several sites is counted several times. Employer establishments only: the self-employed are not counted. Excludes crop and animal production, rail transportation, the postal service, public administration and most government. The source publishes no turnover.
• Business establishments per 1,000 residents (per 1,000 residents). Business establishments with paid employees per 1,000 residents. The two inputs cover different periods (2023 and 2020–2024). Establishments are counted where they operate and residents where they live, so employment centres score high. The self-employed and most government are not counted.
• Residential electricity price (cents per kWh). Average price of electricity to residential customers, by month: revenue from residential sales divided by the kilowatt-hours sold (Form EIA-861M). Average revenue per kilowatt-hour sold, not a tariff: no household pays exactly this rate. Monthly figures come from a sample of utilities, are preliminary until the annual survey replaces them, and are not adjusted for inflation or season.
• Residential natural gas price (USD per thousand cubic feet). Average price of natural gas delivered to residential consumers, by month, in U.S. dollars per thousand cubic feet, taxes and delivery included. Covers gas sold by utilities to their own customers; gas bought from third-party suppliers is excluded except marketer sales in Georgia, New York and Ohio. Some states and months are not available. Not adjusted for inflation or season.

Updates

Each source is reloaded from the publisher’s files when a new release appears; the sources page shows the period loaded and the date the files were retrieved. The whole database can be rebuilt from the raw downloads with one command, and every downloaded file is recorded with its checksum.