City Data Methodology and Coverage

The PEP place universe is mostly incorporated municipalities. Source exceptions include Washington, D.C., consolidated-government balance records, and Urban Honolulu CDP in Hawaii. Urban Honolulu is a census-designated place, not an incorporated city. We preserve these entity labels and do not substitute county-wide or metropolitan totals. Other CDPs and minor civil divisions await separate coverage.

Rankings and original article analysis

Each ranking defines its own population threshold and measurement. Survey rankings require positive exact point estimates with a numerical 90% MOE no greater than half the estimate. The work-from-home ranking also requires 50,000 residents and 5,000 surveyed workers. Ties use competition ranking. Filters retain the original eligible-universe rank. Open-ended medians are listed separately, and no exact rank is assigned to them.

The articles analyze the same exact-GEOID place records. Their exploratory tables generally retain available estimates without the ranking precision screen; the work-from-home article uses its ranking's worker and precision rules. Each article states its minimum population, measure definitions and sources. Charts use exact values and show available MOEs. A point-estimate sort is not a significance test or a quality-of-life score.

Article rent-burden percentages divide the source numerator by all renter-occupied units, including the not-computed category. Cost-of-living pages also show the denominator excluding not-computed ratios. These percentages differ by construction. Uncertainty for the all-renter percentage uses the documented subset formula; an alternative-denominator point calculation does not inherit that MOE. Population, area and density calculations retain their source inputs and cross-product boundary caveats.

Covered places and data periods

The national directory contains 19,483 PEP place records in the 50 states and District of Columbia from Census Population Estimates Program summary level 162. A city, town or village is kept under its source name and seven-digit Census GEOID: two state digits plus five place digits. Other census-designated places, townships recorded only as minor civil divisions, counties, neighborhoods, ZIP codes, ZCTAs and metropolitan areas are outside this directory’s current entity universe.

All 19,483 covered PEP places have stable city profiles. Survey modules are joined by exact seven-digit GEOID to ACS 2020–2024 tables; 19,476 places have a matching population table. Seven unmatched entities retain population and geography, with the survey gap stated. Demographic subpages require a positive survey population and a household table. Housing-cost subpages require a published household-income median, occupied housing units, and at least one published rent or home-value median. Suppressed medians and zero-denominator shares remain unavailable. These criteria yield 19,455 demographic and 18,468 housing-cost pages; no separate page is generated merely to fill a URL quota.

Bedroom-level housing estimates

Supplemental ACS B25031 values use exact place GEOIDs, ACS 2020–2024 and 2024 dollars. We retain each original estimate and its 90% margin of error separately. Negative sentinel codes, blanks and suppressed values are unavailable. The same gross-rent interval definitions apply to bedroom categories; an open-ended median is never presented as an exact dollar amount. A category is labeled low precision when its MOE exceeds half the absolute point estimate. This is a display rule, not a significance test. Bedroom medians describe existing cash-rent units and are not current quotes.

County and metropolitan matching

Every PEP place is matched to all its SUMLEV 157 county place parts using the original state and place codes, retaining zero-population parts as geographic associations. County names and FIPS codes come from the same Vintage 2025 file. Each county FIPS code is then matched to the July 2023 official CBSA delineation list. Multiple counties and CBSAs are retained; no primary county or metro is guessed. A county without a listed association stays unlisted rather than being assumed outside a statistical area, because geography vintages and county codes can differ. These labels describe context, while all statistical tables keep their original place geography.

Nearby station climate matching

Climate uses NOAA’s 1991–2020 U.S. Climate Normals, corrected version 1.0.1. For each Census internal point, we calculate great-circle distance using a mean Earth radius of 6,371.0088 km and select the nearest qualifying U.S. station at most 25 km away. A qualifying station has all 12 months of MLY-TMAX-NORMAL, MLY-TMIN-NORMAL and MLY-PRCP-NORMAL flagged “S” (standard), valid measurement flags and finite values. Standard normals follow NOAA’s availability rules and can include estimates for missing months.

We retain original source values, flags, years used, coordinates, elevation, station ID and distance. Native customary units are degrees Fahrenheit and inches. The displayed series describes that station, not the city’s area average, a particular address or a forecast. Distance alone does not establish identical climate: elevation, coastal exposure and urban conditions can differ. Places without a qualifying nearby station have no substituted climate values.

Population series and ranks

July 1, 2025 is the reference date of the latest population estimate retained here. All July 2020–2025 history values come from Vintage 2025, published May 14, 2026. The revised April 1, 2020 estimate base is labeled separately and is not substituted for the July 2020 estimate or described as an unrevised decennial census count. New population vintages can revise earlier years, so we replace the entire series together.

Absolute annual change = 2025 estimate − 2024 estimate. Percentage annual change = absolute change ÷ 2024 estimate × 100. When the prior estimate is zero, the percentage is unavailable. National and state population ranks use competition ranking: one plus the number of eligible places with a larger value. Growth ranking requires at least 20,000 residents in both years. Population change can reflect boundary changes and is not a measure of net migration.

Open-ended median ranges

Some ACS medians fall in an open-ended interval. In downloaded files, a value such as 3501 represents gross rent of $3,500 or more; it is not an exact $3,501 estimate. We display “At least” or “Less than” with the source interval endpoint. Examples include income of at least $250,000, home value of at least $2,000,000, and rent below $100. The Census 2024 subject definitions, Appendix A define the intervals.

Comparisons do not calculate a difference or home-value-to-income ratio from these ranges. A numeric filter includes a bounded median only if its whole possible range meets the condition; an “at least $3,500” rent is excluded from a maximum-rent filter because its exact upper value is unknown. Downloads retain original source values and add range qualifiers so the encoding can be audited.

Survey values and uncertainty

Income, housing and demographic measures use ACS 2020–2024 five-year detailed tables, published January 29, 2026. Dollar estimates are in 2024 inflation-adjusted dollars. Five-year estimates pool responses across a period and are not a measurement on a single date. Point estimates and available 90% margins of error are read directly from the official table-based summary files.

Derived counts sum mutually exclusive source categories. Their approximate margin is the square root of the sum of squared category margins. A share is the numerator divided by the stated denominator, multiplied by 100. For a subset share, approximate percentage MOE = 100 ÷ denominator × square root of (numerator MOE squared − proportion squared × denominator MOE squared). If that expression is negative, we use the Census general-ratio fallback with addition in place of subtraction. If a required uncertainty component is unavailable, the derived uncertainty is unavailable.

Category counts, denominator counts and their MOEs are retained. A margin greater than half the corresponding estimate is flagged as low precision. These approximations do not reproduce all survey covariance. Comparison differences use point estimates and do not test statistical significance. Overlapping tables are never added together: Hispanic origin can overlap every race category; households with children overlap household size and type.

Negative Census sentinel values are preserved in raw fields and displayed as unavailable, including suppressed, controlled-uncertainty or inapplicable cells. A controlled population total can have an unavailable MOE without an unavailable estimate. Missing data are not zero. Table B16001 lacks the retained city rows, so language pages use the compatible C16001 city-level summary categories.

Housing and denominators

Median gross rent covers renter-occupied units paying cash rent and includes estimated tenant-paid utilities. Owner-occupied home value reflects survey responses, not a completed-sale or current-listing series. Tenure shares use occupied housing units. Vacancy shares use all housing units. The rent-to-income table uses all cash-rent units, including cases where the ratio is not computed; that denominator is explicitly different from a computed-ratio-only burden share.

Median home value ÷ median household income is a ratio of two distributions, not a matched homebuyer budget. We publish no universal salary requirement, household tax estimate or composite cost index. BEA price parities and HUD rental benchmarks are deferred until geographic mappings are validated.

Property taxes, local jobs and city comparisons

Property taxes are ACS B25103 medians of real estate taxes paid, compared with the state and U.S. A city is described as higher or lower than its state only when the difference exceeds the combined 90% margin of error (the square root of the sum of squared MOEs); otherwise the page says the estimates are not statistically different.

Jobs located in a place are LODES workplace counts summed from census blocks. Jobs per 100 residents divides that count by the July population estimate for the same year; it compares workplaces with residents and is not an employment rate. Industry, earnings and age breakdowns are shown only for places with at least 50 jobs; jobs-per-resident positions require 100 jobs and 500 residents, and job change since 2019 requires 250 jobs in 2019.

City profiles place each measure among the same state’s covered places. Survey measures enter these positions only with a coefficient of variation of 30% or less (MOE ÷ 1.645 ÷ estimate). The percentile counts places below the value plus half of any ties. Takeaways list the measures furthest from the state middle (at least 20 percentile points) that also differ from the state beyond their margin of error. Nearby places use straight-line distance between Census internal points. Demographic and housing detail pages for places under 1,000 residents remain available but are excluded from search indexes because most of their survey estimates are imprecise.

How statistical lookalikes are chosen

Each profile lists up to six places elsewhere with a similar statistical profile. Candidates have between half and twice the place’s July 2025 population and lie more than 60 miles from it, because nearby places are listed separately. Seven measures are compared: median household income, median home value, median gross rent and population density (each on a logarithmic scale), median age, the share of adults 25 and older with a bachelor’s degree or higher, and the homeownership rate. Each measure is converted to a standard score using the spread of all reliable place estimates nationwide. A measure counts only when both places have a coefficient of variation of 30% or less, and at least five measures must be shared. The places with the smallest root-mean-square difference are listed.

A lookalike shares measurable Census characteristics; it is not a claim about climate, safety, schools, amenities or how a place feels to live in. Places without five reliable measures show the nearest places by population instead.

Geography, commuting and maps

Approximate density divides the July 2025 place population by 2025 Gazetteer land area. The products can use different within-year boundary dates; density is unsuitable for precise boundary-change work. Land excludes water. Locator maps use Gazetteer internal points and simplified 2024 Census state outlines at 1:20 million. The rectangular longitude/latitude locator is not an area or distance measurement tool.

Employment and commuting figures describe residents. Unemployment divides unemployed civilians by the civilian labor force; participation divides labor-force residents by all people age 16 and older. Travel time excludes people working from home. Vehicle availability describes households. These are separate from jobs located at local establishments or current monthly BLS rates.

Updates, verification and reuse

We retain original selected table rows, labels, source URLs, retrieval and publication dates, checksums, native geography, units and transformations. All displayed measures derive from those shared records. Before a release, we check signs, denominators, category totals, compatible periods and agreement among profiles, charts, filters and comparison data. Corrections trigger a rebuild of affected outputs together.

Visit sources for original files, downloads for observation metadata and updates for the release log. Use the correction form to identify a possible error.