Where every number comes from

BEA Data API (Regional dataset)

Published by U.S. Bureau of Economic Analysis (federal agency).

Kind of number

Model estimate: the same estimates as BEA's downloadable personal income tables, read through BEA's Data API only to check the figures shown

How often

As the underlying tables are released

License

Allowed with credit (BEA API Terms of Service). Credit: U.S. Bureau of Economic Analysis, Data API

Status on this site

Published

Newest figure we show

Used only to check the BEA personal income figures shown on this site against BEA's Data API; it supplies no figure of its own.

Method and terms read

October 15, 2026

How it is made. BEA's Data API serves the same Regional tables as its downloadable files. We read four answers through it (table SQINC1 for the states and the nation, table CAINC1 for the counties; line 1, personal income, and line 3, income per person) and compare every figure with the one in the downloaded file, as text. No figure shown on this site is read from the API: the downloaded files are the source of record and the API is an independent check. Requests are made one after another, four in all, far under BEA's limits.

Who is in it. All data in BEA's Regional Interactive Tables.

Who is not. Where BEA does not publish a figure its downloadable file prints (NA); the API answers the same cell with the text 0. We never read that 0 as a figure: the eight Connecticut counties, whose 2024 figure BEA does not publish (it moved to planning regions), are shown as not available, and the check lists them. The user guide does not describe this.

Revisions. Each API answer carries the same 'Last updated' note as the downloaded file of the same table (state table September 30, 2026; county table February 5, 2026), so the two are the same vintage. If they ever differ, the check stops and neither is chosen until both are read again.

Required notice. This product uses the Bureau of Economic Analysis (BEA) Data API but is not endorsed or certified by BEA.

The provider's method → Its terms →

BEA county personal income, yearly

Published by U.S. Bureau of Economic Analysis (federal agency).

Kind of number

Model estimate: estimates built from many federal sources, by where people live, in current dollars

How often

Yearly, 11 to 13 months behind

License

Public domain (U.S. government statistics). Credit: U.S. Bureau of Economic Analysis

Status on this site

Published

Newest figure we show

Describes 2024

Retrieved

October 15, 2026

Next release

December 2, 2026 (2025)

Method and terms read

October 15, 2026

Used on

Faster signals

How it is made. County personal income is the income of a county's residents, wherever they work, from wages, proprietors' income, dividends, interest, rents and government benefits. Per capita income divides it by the Census Bureau's July 1 population. It is in current dollars, not adjusted for inflation, and is revised at each release.

Who is in it. Counties and county equivalents (parishes, Alaska boroughs and Census areas, DC, Maryland, Missouri, Nevada and Virginia independent cities), states and the nation.

Who is not. Kalawao County, Hawaii and Virginia's small independent cities (fewer than 100,000 residents) are combined with an adjacent county in one area, so they have no separate figure. From 2024 Connecticut is published by its nine planning regions, not its eight counties.

Revisions. Estimates are released and then revised on a regular schedule and to match revisions of the national accounts, so earlier years can change at each release. The downloaded file carries one vintage, named in its last note row. Our comparison uses the same file for both years.

Required notice. This product uses the Bureau of Economic Analysis (BEA) Data API but is not endorsed or certified by BEA.

The provider's method → Its terms →

BEA state personal income, quarterly

Published by U.S. Bureau of Economic Analysis (federal agency).

Kind of number

Model estimate: estimates built from many federal sources, seasonally adjusted at annual rates, in current dollars

How often

Quarterly, about three months behind

License

Public domain (U.S. government statistics). Credit: U.S. Bureau of Economic Analysis

Status on this site

Published

Newest figure we show

Describes 2026 Q2

Retrieved

October 15, 2026

Next release

December 23, 2026 (2026 Q3)

Method and terms read

October 15, 2026

Used on

Faster signals

How it is made. Personal income is the income persons receive from all sources: labor, owning a home or business, financial assets, and transfers. BEA estimates it for states quarterly, seasonally adjusts it and states it at annual rates (the amount that would be registered if the quarter's rate were kept for a full year). Per capita income divides it by mid-quarter population. It is in current dollars, not adjusted for inflation.

Who is in it. The United States, 50 states and the District of Columbia, and BEA's eight regions.

Who is not. No county detail in this table; counties have their own annual table.

Revisions. Quarterly estimates are revised on a schedule (the quarters of earlier years are revised at each annual update) and for new seasonal factors, so the whole series can change at each release. Our comparison uses the same file for both quarters.

Required notice. This product uses the Bureau of Economic Analysis (BEA) Data API but is not endorsed or certified by BEA.

The provider's method → Its terms →

Consumer Price Index

Published by U.S. Bureau of Labor Statistics (federal agency).

Kind of number

Survey sample

How often

Monthly

License

Public domain (U.S. government statistics). Credit: U.S. Bureau of Labor Statistics, Consumer Price Index

Status on this site

Research preview

Newest figure we show

Describes August 2026

Retrieved

September 28, 2026

Method and terms read

Not readable from our environment now; our data package recorded the documentation it used on October 12, 2026

Used on

Economic score, Map and county profiles

How it is made. Not read: no BLS CPI method document could be fetched in this registry.

Who is in it. Not read from BLS. Project record: 23 metro areas publish a monthly food-at-home index, plus Census regions and the U.S. city average; counties outside those metros get a regional or national benchmark (the site labels this).

Who is not. Not read from BLS. Counties are not CPI geographies.

Revisions. Not read from BLS.

The provider's method → Its terms →

Local Area Unemployment Statistics

Published by U.S. Bureau of Labor Statistics (federal agency).

Kind of number

Model estimate: model-based estimates that combine a household survey, payroll survey and unemployment-insurance counts

How often

Monthly, about seven weeks after the month

License

Public domain (U.S. government statistics). Credit: U.S. Bureau of Labor Statistics, Local Area Unemployment Statistics

Status on this site

Published

Newest figure we show

Describes August 2026

Retrieved

October 8, 2026

Method and terms read

Not readable from our environment now; our data package recorded the documentation it used on October 8, 2026

Used on

Economic score, Map and county profiles

How it is made. Not read from BLS: its website refuses automated readers, so no BLS method page could be fetched for this registry. Our own data packages describe the series as monthly county estimates, not seasonally adjusted, by place of residence, with the 2025 redesign re-estimating the history; those statements are ours, not quotations from BLS.

Who is in it. Not read from BLS.

Who is not. Not read from BLS. Project record: Connecticut's eight historical counties have no series (BLS replaced them with nine planning regions from January 2025 data).

Revisions. Not read from BLS. Project record: annual revision each spring (2026-05-19 revised 2016-2025 substate data); the newest month is preliminary.

The provider's method → Its terms →

American Community Survey, 5-year estimates

Published by U.S. Census Bureau (federal agency).

Kind of number

Survey sample

How often

Yearly; each release pools five years of interviews

License

Allowed with credit (U.S. government statistics). Credit: U.S. Census Bureau, American Community Survey 5-year estimates

Status on this site

Published

Newest figure we show

Describes 2020-2024 (a five-year average)

Retrieved

October 14, 2026

Method and terms read

October 15, 2026

Used on

District and county context, Economic score, Map and county profiles, Young-adult context

Editions used. 2015-2019 (map demographics and the 2020 model) and 2020-2024 (economic score, young-adult context, district profiles)

How it is made. A continuous address-based sample survey (internet, mail, then personal interview for a subsample of nonrespondents), including group quarters. A 5-year estimate is a period estimate: the average over the 60 months, not the value in any one year. Weights are controlled to an average of the Population Estimates Program's county estimates over the period. Dollar values are restated to the last year of the period with the national CPI. Every published estimate carries a 90% margin of error.

Who is in it. Households and group-quarters residents (barracks, dorms, nursing homes, prisons) sampled from every county and county-equivalent; 5-year estimates exist for all areas.

Who is not. No one is excluded by design, but small areas rest on small samples (large margins of error), and 5-year periods containing 2020 have larger margins because of fewer pandemic-year interviews. Some data are modified or suppressed for confidentiality (data swapping, synthetic data).

Revisions. The documents read do not describe revisions of a published 5-year release; each release is a new, separate period estimate. Corrections are announced on the ACS errata page (linked from the technical documentation page, not read). Population-estimate controls use the vintage of the last year of the period, so two 5-year releases rest on different PopEst vintages.

Required notice. This product uses the Census Bureau Data API but is not endorsed or certified by the Census Bureau.

The provider's method → Its terms →

Business Formation Statistics

Published by U.S. Census Bureau (federal agency).

Kind of number

Administrative records: applications for an employer identification number (EIN) filed with the IRS

How often

Monthly, about two weeks after the month

License

Allowed with credit (U.S. government statistics). Credit: U.S. Census Bureau, Business Formation Statistics

Status on this site

Published

Newest figure we show

Describes September 2026

Retrieved

October 15, 2026

Next release

November 12, 2026 (October data, with the annual update)

Method and terms read

October 15, 2026

Used on

Faster signals

How it is made. BFS counts applications for an Employer Identification Number on IRS Form SS-4, filtered to applications made predominantly for business purposes (BA). High-propensity applications (HBA) are the subset likely to become employers (corporations, planned wages, hiring, certain industries). The formation series follow applications forward to the first payroll filing and are partly projected by a model; they are monthly and national/state only. An application is not a new business: many never hire anyone.

Who is in it. EIN applications made predominantly for business purposes in the 50 states and DC (state series), plus Puerto Rico in the monthly application series. Includes applications that will never become employers (BA); HBA narrows to likely employers.

Who is not. Businesses started without an EIN (many sole proprietors who use a Social Security number), applications from public entities, tax liens/estates/trusts and certain financial filings, applications in some low-transition agricultural, financial and household industries, and applications with no state information. Formation series exclude new employers formed by existing firms and non-employer businesses. Puerto Rico and Island Areas are excluded from national and regional totals.

Revisions. Each month the seasonally adjusted series are revised for the two prior months and for the current and previous month a year earlier. Once a year (next: the October 2026 data, released 2026-11-12) seasonal factors are recomputed, NAICS codes for the previous five years are revised, and formation series are updated. The documents do not describe revisions of the weekly not-seasonally-adjusted counts; weekly values are rounded and do not sum to the monthly estimates.

The provider's method → Its terms →

Building Permits Survey, annual county estimates

Published by U.S. Census Bureau (federal agency).

Kind of number

Survey sample: a survey of permit-issuing places, with imputation for places that do not report

How often

Yearly, about five months after the year

License

Allowed with credit (U.S. government statistics). Credit: U.S. Census Bureau, Building Permits Survey

Status on this site

Research preview

Newest figure we show

Describes 2025

Retrieved

September 28, 2026

Method and terms read

October 15, 2026

Used on

Map and county profiles

How it is made. The annual county file is a full tabulation of all permit offices in the universe, not a sample. For each place the annual figure is, in order: the place's reported annual total, the sum of its 12 monthly reports, or an imputed annual value. Imputation for non-reporters multiplies the place's prior-year total by the ratio of current to prior activity among reporting places in the same cell. The file carries two versions of every figure: estimates with imputation, and reported only.

Who is in it. New privately-owned residential housing units authorized by permit-issuing places; over 99% of private residential construction is in permit-issuing places.

Who is not. Areas that need no permit, publicly-owned construction (the survey covers privately-owned units), group quarters, hotels/motels, manufactured (mobile) homes, moved buildings and units created inside existing structures. Connecticut is reported by planning region.

Revisions. Annual figures replace monthly sums for each place where an annual report exists; after annual processing only the U.S. and regional monthly totals are benchmarked to the annual totals. The documents read do not describe later revisions to a published annual county file. Since January 2023 the universe of permit offices is updated every January.

The provider's method → Its terms →

Building Permits Survey, monthly county files

Published by U.S. Census Bureau (federal agency).

Kind of number

Survey sample: a survey of permit-issuing places, with imputation for places that do not report

How often

Monthly, about seven weeks after the month

License

Allowed with credit (U.S. government statistics). Credit: U.S. Census Bureau, Building Permits Survey

Status on this site

Published

Newest figure we show

Describes January to August 2026

Retrieved

October 15, 2026

Method and terms read

October 15, 2026

Used on

Faster signals

How it is made. The survey covers all places that issue building or zoning permits for new privately owned housing. Since January 2022 slightly under half of the roughly 20,000 permit-issuing places (those averaging 6 or more permits a year; about 8,400 places) are asked to report monthly, and the rest only annually. Monthly county figures since January 2022 combine reported data with imputed values for non-respondents and for places not in the monthly survey; each file gives both an 'estimates with imputation' set and a 'reported only' set. Imputation scales each place's prior annual total by the current-month ratio of reporting places in the same imputation cell.

Who is in it. New privately owned housing units authorized by building or zoning permits in permit-issuing places, by structure size. Since January 2022 the monthly county file lists every county in the current universe (3,019 county rows in co2608c.txt), with imputed values where places did not report or are surveyed only annually.

Who is not. Units built where no permit is required (under 1 percent nationally, but varying a lot by state), publicly owned housing, HUD-code manufactured homes, group quarters, transient accommodations, and units created inside existing buildings. Permits are authorizations, not starts or completions. Before January 2022, monthly county totals included only places in the monthly sample, so counties not fully covered were incomplete; the December 2021 file has only 750 county rows (our count), which is where the blanks in older monthly files come from.

Revisions. Monthly county files (c) are not revised after publication. Late reports and corrections go into the year-to-date (y) files, so summing monthly c files does not reproduce the year-to-date total, and the annual file (reported annual data first, then the sum of 12 monthly reports, then annual imputation) is the final county figure. Only U.S. and regional monthly totals are benchmarked to the annual totals. The universe of permit-issuing places is updated each January. In 2024 the unit response rate for monthly estimates was about 66 percent (81 percent after the annual survey).

The provider's method → Its terms →

TIGER/Line and cartographic boundary files, congressional districts

Published by U.S. Census Bureau (federal agency).

Kind of number

Administrative records: official boundaries

How often

Yearly, and for each Congress

License

Allowed with credit (U.S. government data). Credit: U.S. Census Bureau, TIGER/Line and cartographic boundary files

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Elections, Redistricting

How it is made. TIGER/Line Shapefiles are extracts of the MAF/TIGER database with legal boundaries as of January 1 of the file year, at full detail. Cartographic boundary files are simplified (generalized) versions designed for small-scale thematic maps; they drop small holes and parts and are clipped to a simplified shoreline, so they must not be used for area calculation, geocoding or exact area relationships. Neither product contains demographic data.

Who is in it. All congressional districts (and counties, states) of the U.S. as of the file's boundary date.

Who is not. Cartographic boundary files omit some small areas and offshore parts and may not align with the same areas in another year's file.

Revisions. Each year's file is a new snapshot; boundaries are statistical depictions, not legal land descriptions, and the Census Bureau gives no warranty of positional accuracy.

The provider's method → Its terms →

Population Estimates, county population and components of change

Published by U.S. Census Bureau (federal agency).

Kind of number

Model estimate: estimates built from births, deaths and migration records

How often

Yearly vintage, each revising the series back to 2020

License

Allowed with credit (U.S. government statistics). Credit: U.S. Census Bureau, Population Estimates Program

Status on this site

Published

Newest figure we show

Describes July 1, 2025 (V2025)

Method and terms read

October 15, 2026

Used on

Map and county profiles

How it is made. A cohort-component model: start from the 2020 Census base (the MARC file, adjusted for the young-child undercount), add births, subtract deaths, add net domestic and international migration, then rake to consistent totals. County net domestic migration comes from IRS tax returns (ages 0-64), Medicare enrollment (65+), SSA records and group-quarters change. Raking leaves a residual that is not a demographic component. Vintage 2025 had two input gaps: no new CMS data for ages 65+ (prior-year rates reused) and only 26 weeks of 2025 IRS data.

Who is in it. The resident population of every county and county-equivalent; Connecticut is estimated by its nine planning regions (not the eight historical counties).

Who is not. No sampling; but the numbers are modeled, so domestic migration reflects only who appears in tax, Medicare and SSA records. Connecticut's historical counties are not in Vintage 2025.

Revisions. Every annual vintage revises the entire series back to April 1, 2020 and supersedes the previous vintage, so values from different vintages must not be mixed. Vintage 2025 also changed the base population method substantially (full MARC detail).

The provider's method → Its terms →

Quarterly Workforce Indicators

Published by U.S. Census Bureau (LEHD program) (federal agency).

Kind of number

Administrative records: unemployment-insurance wage records and other administrative data, noise-infused for confidentiality

How often

Quarterly, about three quarters behind

License

Allowed with credit (U.S. government statistics). Credit: U.S. Census Bureau, Quarterly Workforce Indicators

Status on this site

Published

Newest figure we show

Describes jobs on October 1, 2025 (the first day of 2025 Q4)

Retrieved

October 14, 2026

Method and terms read

October 15, 2026

Used on

Young-adult context

How it is made. QWI are built from job histories in the LEHD linked employer-employee data: each quarter's UI wage record ties a worker to an employer, and measures count jobs (a person with two employers counts twice) at the employer's place of work. Missing information is imputed with random draws, and every released number is protected with multiplicative noise ('fuzzed'). The data are administrative, so there is no sampling error, but there is nonsampling error.

Who is in it. Jobs covered by state unemployment insurance (about 96% of wage and salary civilian jobs), private and state/local government.

Who is not. Federal government jobs (not yet in regular releases), self-employment, railroad and some agricultural jobs. States whose series are not current: Alaska (data end 2016Q2) and Michigan (2021Q4) in R2026Q3.

Revisions. The whole QWI time series is recalculated at every quarterly release (new state submissions, algorithm changes, new imputation draws), and the Census Bureau advises not to combine data from different releases. A refresh must re-pull every quarter used.

Required notice. This product uses the Census Bureau Data API but is not endorsed or certified by the Census Bureau.

The provider's method → Its terms →

Congress.gov API, member records

Published by Library of Congress (federal agency).

Kind of number

Administrative records

How often

Continuous

License

Terms unclear (U.S. government data; no API license stated). Credit: Congress.gov, Library of Congress

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

District and county context

How it is made. A REST API (XML or JSON, version 3) over Congress.gov collections. Congress.gov republishes what the House and Senate deliver; suspected errors are referred back to the chambers. An api.data.gov key is required; 5,000 requests per hour.

Who is in it. Members of Congress (current and former), bills, amendments, committees and other collections.

Who is not. Vacancies are documented by the House Clerk, not this API.

Revisions. API changes are announced in the ChangeLog; data errors are triaged and referred to the chambers.

The provider's method → Its terms →

Low-Income Energy Affordability Data (LEAD) Tool

Published by U.S. Department of Energy (federal agency).

Kind of number

Model estimate: modeled energy costs, not observed bills

How often

Irregular editions

License

Allowed with credit (U.S. government data). Credit: U.S. Department of Energy, LEAD Tool

Status on this site

Research preview

Newest figure we show

Describes 2018-2022 (American Community Survey microdata)

Retrieved

September 28, 2026

Method and terms read

October 15, 2026

Used on

Economic score, Map and county profiles

How it is made. LEAD starts from households' self-reported monthly electricity, gas and other-fuel spending in ACS microdata, annualizes it, adjusts for inflation and calibrates it to EIA utility revenue by service territory. PUMS records are assigned to census tracts and aggregated up. Energy burden is average annual housing energy cost divided by average household income (closer to mean than median income), rounded to whole percent. It is a modeled estimate, not observed bills, and excludes transportation.

Who is in it. Households in the 50 states, DC and Puerto Rico, at state, county, city, census tract and tribal-area levels.

Who is not. Transportation energy; household counts are LEAD's own estimates and do not exactly match ACS; small samples can inflate tract-level burdens.

Revisions. Updates are new releases built on a newer ACS window (the current one is 2018-2022); the documents read do not describe revisions within a release or a schedule for the next update.

The provider's method →

Unemployment insurance weekly claims, by state (ETA 539)

Published by U.S. Department of Labor, Employment and Training Administration (federal agency). The underlying data come from State unemployment insurance agencies (state or local government).

Kind of number

Administrative records: counts of claims filed with state unemployment insurance programs

How often

Weekly; the state week is one week behind the Thursday release

License

Public domain (U.S. government work, public domain). Credit: U.S. Department of Labor, Employment and Training Administration

Status on this site

Published

Newest figure we show

Describes week ending October 3, 2026

Retrieved

October 15, 2026

Next release

Every Thursday at 8:30 a.m. Eastern (the state week in the file follows one week later)

Method and terms read

October 15, 2026

Used on

Faster signals

How it is made. States report weekly counts of UI claims to DOL. The advance figures in the Thursday release come from the ETA 538 (claims taken by the liable state); the following week the state figures are replaced by ETA 539 counts by state of residence (agent state), and the downloadable state file is the ETA 539 series. Initial claims count new claims after a separation; continued weeks claimed count weeks of benefits claimed (not unique people) and are DOL's measure of insured unemployment. States must adjust counts for administrative (non-economic) distortions before reporting; that is not seasonal adjustment.

Who is in it. Claims under each state's regular UI program (c3, c8), with UCFE (former federal civilian employees), UCX (ex-servicemembers), workshare, Extended Benefits and state additional-benefit claims in separate columns. Covered employment includes private, state and local government and nonprofit employment covered by state law. All 50 states, DC, Puerto Rico and the Virgin Islands.

Who is not. Unemployed people who do not file or are not eligible for UI; federal government employment is excluded from covered employment. Continued weeks claimed are not a count of unique individuals. No county or metro geography. Advance-week state figures are by liable state, revised figures by state of residence, so the two are not directly comparable.

Revisions. Advance figures (ETA 538) are replaced the next week by ETA 539 figures by state of residence; over-the-week changes in the release reflect any prior-week revisions states submit. Each January, with BLS's new seasonal factors, ETA revises the historical claims series for updates to the unadjusted data. Some state figures can be OUI estimates (Puerto Rico and South Carolina were marked as estimates in the 2026-10-15 release). Because ar539.csv is rewritten each week, older weeks can change; snapshot the file at each acquisition.

The provider's method → Its terms →

Form EIA-860M, Monthly Update to the Annual Electric Generator Report

Published by U.S. Energy Information Administration (federal agency). We download it from Catalyst Cooperative (PUDL) (university or nonprofit).

Kind of number

Survey sample

How often

Monthly

License

Public domain (U.S. government statistics). Credit: U.S. Energy Information Administration, Form EIA-860M, via PUDL

Status on this site

Published

Newest figure we show

Describes July 2026

Method and terms read

October 15, 2026

Used on

Data centers

How it is made. EIA's monthly estimate of existing and proposed generating units (1 MW and up), combining the monthly survey with annual EIA-860 data that may not yet be verified. Capacities are preliminary and are corrected in later months without notice; they are not capacity commitments. The final inventory is the annual EIA-860.

Who is in it. Existing, proposed and (since March 2017) retired generators at plants with at least 1 MW combined nameplate capacity, including Puerto Rico since March 2018.

Who is not. Plants under 1 MW.

Revisions. Each month's inventory may silently correct earlier estimates; the annual EIA-860 is the final inventory.

The provider's method → Its terms →

Form EIA-861, Annual Electric Power Industry Report

Published by U.S. Energy Information Administration (federal agency). We download it from Catalyst Cooperative (PUDL) (university or nonprofit).

Kind of number

Survey sample

How often

Yearly

License

Public domain (U.S. government statistics). Credit: U.S. Energy Information Administration, Form EIA-861, via PUDL

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Data centers

How it is made. A census of all U.S. electric utilities: each utility reports annual sales, revenue and customers by state and sector, operational data, and the counties it serves (the peak-demand fields the site uses were not described on the page read). Smaller bundled utilities file the short form EIA-861S, and EIA adjusts state totals for non-respondents. Service territory lists counties where a utility has equipment, not the share of load in each county.

Who is in it. All U.S. electric utilities and power marketers selling to end users.

Who is not. Short-form (EIA-861S) respondents give less detail and since 2020 some of their values are imputed or estimated.

Revisions. Each data year appears first as early-release data and later as final data. The site's package read PUDL's copy when the 2025 year was provisional; EIA released final 2025 data on 2026-10-07.

The provider's method → Its terms →

Electric Power Monthly, average residential price (Table 5.6.A)

Published by U.S. Energy Information Administration (federal agency).

Kind of number

Survey sample

How often

Monthly, about two months behind

License

Public domain (U.S. government statistics). Credit: U.S. Energy Information Administration, Electric Power Monthly

Status on this site

Research preview

Newest figure we show

Describes July 2026

Retrieved

September 28, 2026

Method and terms read

October 15, 2026

Used on

Economic score, Map and county profiles

How it is made. EIA collects monthly sales and revenue from a cutoff sample of utilities and uses a regression (ratio) model to estimate the rest of the universe, with the annual EIA-861 census as the frame. State average price is estimated revenue divided by estimated sales by sector, so it is a weighted average across all customers and rate schedules, including taxes and fees, not any customer's tariff rate. Monthly values are preliminary estimates.

Who is in it. Sales to ultimate customers by all U.S. electric utilities and energy service providers (sample plus model), by state and sector.

Who is not. No sub-state detail: one value per state, so every county in a state gets the same figure on the site.

Revisions. All monthly values are first preliminary; they are revised once the prior year's annual data are final, and otherwise only for significant errors. The year-ago column therefore can differ from what was published a year earlier.

The provider's method → Its terms →

EIA weekly retail gasoline prices, regular (Gasoline and Diesel Fuel Update)

Published by U.S. Energy Information Administration (federal agency).

Kind of number

Survey sample: cash pump prices from a stratified sample of about 1,000 retail outlets, weighted by sales volume

How often

Weekly (Monday prices)

License

Public domain (U.S. government statistics). Credit: U.S. Energy Information Administration

Status on this site

Published

Newest figure we show

Describes the week of October 12, 2026

Retrieved

October 15, 2026

Next release

October 20, 2026

Method and terms read

October 15, 2026

Used on

Faster signals

How it is made. Every Monday EIA collects the cash pump price for regular, midgrade and premium gasoline from a stratified sample of about 1,000 retail outlets (Form EIA-878). Prices include taxes and are the 8:00 a.m. self-serve price. Missing prices are imputed. EIA weights the prices by outlet sales volume to estimate averages for the nation, regions and selected cities and states.

Who is in it. The United States, EIA's regions (the Petroleum Administration for Defense Districts, PADD 1 with its three sub-districts 1A, 1B and 1C, and PADDs 2 to 5) and nine states: California, Colorado, Florida, Massachusetts, Minnesota, New York, Ohio, Texas and Washington. EIA also publishes ten cities.

Who is not. No county figures and no figures for the other 41 states and DC (their region's figure is the nearest EIA publishes). EIA says it has no county prices.

Revisions. EIA does not revise published weekly prices: when it found formulation misclassifications in 2023 it said it would not change them. A new sample on May 14, 2018 broke comparability with earlier weeks, so a comparison with a week before that date is not valid (ours reach back only one year).

The provider's method → Its terms →

EIA average residential electricity price by state (retail sales)

Published by U.S. Energy Information Administration (federal agency).

Kind of number

Survey sample: revenue divided by sales, from a sample of utilities plus a model; preliminary

How often

Monthly, about two months behind

License

Public domain (U.S. government statistics). Credit: U.S. Energy Information Administration

Status on this site

Published

Newest figure we show

Describes July 2026

Retrieved

October 15, 2026

Next release

October 23, 2026

Method and terms read

October 15, 2026

Used on

Faster signals

How it is made. EIA estimates monthly sales and revenue by state and sector from a sample of utilities plus a model, then divides revenue by sales. The price is a weighted average over all residential customers and rate schedules, not any customer's tariff. Monthly values are preliminary.

Who is in it. One residential price per state and the District of Columbia, plus a U.S. total, every month.

Who is not. No sub-state detail: one value per state, so it is never a county's price.

Revisions. Monthly values are first preliminary and are revised after the prior year's data are final, so the same month a year earlier can differ from what was published then; our comparison uses what the API serves now for both months.

The provider's method → Its terms →

Federal Elections (official results compilations)

Published by Federal Election Commission (federal agency). The underlying data come from State election officials' certified results (state or local government).

Kind of number

Compilation

How often

After each federal election

License

Public domain (U.S. government publication). Credit: Federal Election Commission, Federal Elections

Status on this site

Published

Newest figure we show

Describes elections 2012, 2016 and 2020

Method and terms read

October 15, 2026

Used on

Findings

How it is made. The FEC compiles federal election results as officially reported and certified by state and territory election offices; it does not administer or certify elections. Federal Elections is a biennial compilation of primary, runoff and general results for Senate and House (and President in presidential years). Totals are valid votes for candidates; blank, void, under and over votes are excluded; percentages are re-rounded uniformly and may not match state figures.

Who is in it. Senate, House, Delegate and (in presidential years) President; primary, runoff and general elections, plus special elections in an appendix.

Who is not. County-level results are not in these publications (state and district only). 2024 congressional results not yet published (only presidential results and a candidates-on-ballot list).

Revisions. Errata supplements are issued if states amend results.

The provider's method → Its terms →

FHFA House Price Index

Published by Federal Housing Finance Agency (federal agency). The underlying data come from Fannie Mae and Freddie Mac mortgage records (private company).

Kind of number

Administrative records: repeat sales of the same homes, from mortgages bought or guaranteed by Fannie Mae and Freddie Mac

How often

Quarterly for states; yearly for counties

License

Public domain (U.S. government work, generally public domain). Credit: Federal Housing Finance Agency, House Price Index

Status on this site

Published

Newest figure we show

Describes 2026 Q2

Retrieved

October 15, 2026

Next release

October 27, 2026 (monthly); November 24, 2026 (quarterly, 2026 Q3)

Method and terms read

October 15, 2026

Used on

Faster signals

How it is made. A weighted repeat-sales index (a modified Case-Shiller method) built from repeat transactions on the same single-family properties whose mortgages Fannie Mae or Freddie Mac bought or securitized since 1975. The purchase-only index uses sale prices only; the all-transactions index adds appraisal values from refinance mortgages. Division and national indexes are built up from state indexes weighted by each state's share of one-unit detached housing. Indexes are nominal (not inflation-adjusted) and measure price changes, not price levels.

Who is in it. Single-family properties with conforming, conventional mortgages bought or securitized by Fannie Mae or Freddie Mac (purchase-only: sale prices; all-transactions: sale prices plus refinance appraisals). States and DC, every MSA (metro divisions where they exist); purchase-only metro indexes only for the 100 largest. County, ZIP and tract indexes are annual and developmental.

Who is not. Non-conforming (e.g., jumbo) and non-conventional (FHA, VA) mortgages, cash sales (except in the Expanded-Data index), condominiums, cooperatives, multi-unit properties and planned unit developments. Small counties with thin samples are not reported or carry missing values. Eagle Pass, TX has no metro index.

Revisions. Every release revises the whole history: new repeat transactions, late-acquired (seasoned) loans and the roughly two-month data lag all change earlier index values, most of all for recent periods. A stored figure must carry the release (vintage) it came from.

Required notice. This product uses FHFA data but is neither endorsed nor certified by FHFA.

The provider's method → Its terms →

Office of the Clerk, U.S. House: member data and election statistics

Published by Office of the Clerk, U.S. House of Representatives (federal agency).

Kind of number

Administrative records

How often

When membership changes; election statistics after each election

License

Public domain (U.S. government work, public domain). Credit: Office of the Clerk, U.S. House of Representatives

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Elections, Redistricting

How it is made. The member list is compiled by the Clerk and holds current members only, with committee assignments. The election statistics compile, from official sources, the vote for presidential electors and for each Senate, House, Resident Commissioner and Delegate nominee, with a recapitulation by state.

Who is in it. Current House members (member data); every federal nominee's vote by state and district (statistics).

Who is not. Historic members are largely absent from MemberData.xml.

Revisions. MemberData.xml is reissued with a new publish-date on each update.

The provider's method → Its terms →

SOI county-to-county migration data

Published by Internal Revenue Service, Statistics of Income (federal agency).

Kind of number

Administrative records: addresses on individual income-tax returns in two consecutive years

How often

Yearly, about two years behind

License

Public domain (U.S. government statistics). Credit: Internal Revenue Service, Statistics of Income, county-to-county migration data

Status on this site

Published

Newest figure we show

Describes returns filed in 2022and2023

Method and terms read

October 15, 2026

Used on

Map and county profiles

How it is made. IRS matches individual returns filed in two consecutive calendar years by taxpayer ID (primary, secondary and dependent filers) and compares the county of the mailing address on each. A return whose county changed is a migrant; individuals are the people on the return, and AGI comes from the year-2 return. Returns are placed in counties with a Census ZIP+4-to-county crosswalk. From 2022-2023 the matching method changed, adding about 5 percent more returns, so 2022-2023 is a new series.

Who is in it. Filers (and the people on their returns) who filed in both years with matching taxpayer IDs; includes ITIN filers.

Who is not. People not required to file, filers who missed a year or switched ITIN to SSN, people claimed as dependents in year 2, returns without a ZIP code. County flows under 20 returns are suppressed or pooled into 'Other Flows', and dominant returns are removed, so county totals need not add to state totals.

Revisions. The documents do not describe revisions to a released pair; each pair is a separate release. Comparisons across years are affected by the annual ZIP-to-county crosswalk revisions (absorbed in the non-migrant counts), changes in return processing, and the 2022-2023 series break.

The provider's method → Its terms →

OpenFEC API, candidates and campaign finance

Published by Federal Election Commission (federal agency).

Kind of number

Administrative records: reports campaigns file with the FEC

How often

Nightly

License

Allowed with credit (OpenFEC API terms of service). Credit: Source: openFEC API (Federal Election Commission)

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

District and county context

How it is made. A REST API over the FEC's campaign finance data, tied to the filed forms by file and image IDs and updated nightly. A key from api.data.gov allows 1,000 calls an hour, 100 results per page.

Who is in it. Candidates and committees filing with the FEC.

Who is not. Not stated in the documents read.

Revisions. Not stated in the documents read (data are updated nightly).

The provider's method → Its terms →

U.S. Senate, senators' contact list

Published by United States Senate (federal agency).

Kind of number

Administrative records

How often

When membership changes

License

Terms unclear (U.S. government data; no reuse terms stated). Credit: United States Senate

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Elections

How it is made. The Senate's own contact list of sitting senators, published as a web page with a PDF and an XML version. No method statement; it is an official roster.

Who is in it. The 100 sitting senators.

Who is not. Former senators; vacancies are not listed as records.

Revisions. Overwritten in place; last_updated shows the latest change.

The provider's method → Its terms →

The White House, Administration page

Published by The White House (federal agency).

Kind of number

Administrative records

How often

When officeholders change

License

Public domain (U.S. government work, not copyrighted). Credit: The White House

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Elections

How it is made. The White House's own page listing the President and Vice President. The site uses it only to name the current President on the presidential seat page.

Who is in it. The current President and Vice President.

Who is not. Anything else; the site takes no figures from it.

Revisions. Updated when officeholders change.

The provider's method → Its terms →

State and county election officials' candidate lists and certifications

Published by State and county election officials (state or local government).

Kind of number

Administrative records

How often

Each election

License

Terms unclear (Public records; terms not read). Credit: Each document's own publisher, named beside it

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

Not readable from our environment now; our data package recorded the documentation it used on October 15, 2026

Used on

Elections

How it is made. By title, the documents are candidate lists and ballot certifications, not results: 34 of 49 titles contain 'candidate', 15 'official', 6 'certified' and 4 'ballot' (e.g. California 'Certified List of Candidates for the November 3, 2026, General Election'; Texas 'Ballot Certification Report'; Ohio SOS 'Directive 2026-45 - Form of the Official Ballot'; New Jersey 'Official List - Candidates for House of Representatives'; Georgia's certified write-in list posted by Fayette County). Several county lists are labelled 'unofficial' (e.g. Berrien and Genesee counties, Michigan). No certified canvass of results is in the manifest.

Who is in it. Candidates for U.S. Senate and House on 2026 general-election ballots in the states and counties listed.

Who is not. Most states are not represented by a state-level document (21 states plus Michigan/Ohio counties); no election results.

Revisions. Candidate lists are dated snapshots (several titles carry 'updated' or 'as of' dates); later withdrawals or replacements would need a new retrieval.

The provider's method →

50-State Redistricting Simulations

Published by The ALARM Project (university or nonprofit).

Kind of number

Model estimate: simulated district plans

How often

Versioned releases

License

Public domain (CC0 (data)). Credit: McCartan et al., 50-State Redistricting Simulations, Harvard Dataverse

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Redistricting

How it is made. For each state the project samples 5,000 alternative congressional plans with the redist simulation software, under the project's best approximation of the state's rules (contiguity, population equality within 0.5% unless the state sets a limit, compactness, county-split limits, and in some states VRA targets), and scores the enacted plan and the simulations on the same measures (expected seats from averaged 2016-2020 statewide races, efficiency gap, partisan symmetry, Polsby-Popper, counties split). The authors say the simulations are not a legal evaluation.

Who is in it. Congressional plans of states that enacted a multi-district map and were analyzed (45 states listed as analyzed on the project page); 2010 and 2020 cycles on Dataverse.

Who is not. Single-district states, states not yet analyzed, and (per the FAQ) state legislative plans except where added later.

Revisions. Dataverse versions replace earlier ones (version 15 seen); the dataset notes point to a change log on GitHub. A stored file must keep its Dataverse version and file id.

The provider's method → Its terms →

United States Congressional District Shapefiles

Published by Jeffrey B. Lewis, Brandon DeVine, Lincoln Pitcher and Kenneth C. Martis (university or nonprofit).

Kind of number

Compilation: digitized historical district boundaries

How often

Irregular versions

License

Allowed with credit (MIT License). Credit: Lewis, DeVine, Pitcher and Martis, Digital Boundary Definitions of United States Congressional Districts

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Redistricting

How it is made. District shapes for every Congress were built by aggregating historical county shapes (NHGIS, Newberry Atlas) where district lines followed counties, and drawn district by district from legal and cartographic sources elsewhere, relying heavily on Martis' Historical Atlas. Later boundaries come from the Census Bureau or state agencies. Since version 2.0 the GeoJSON files are the canonical, full-resolution form.

Who is in it. Every U.S. congressional district used between 1789 and 2025.

Who is not. Detailed construction notes for individual districts are available only on request.

Revisions. Maintained with a patching workflow; the project pins commit aba2b6b, whose README and LICENSE are identical to the current master copies read on 2026-10-15.

The provider's method → Its terms →

All About Redistricting

Published by Justin Levitt (Loyola Law School) (university or nonprofit).

Kind of number

Compilation

How often

Continuous

License

Terms unclear (No license stated). Credit: Justin Levitt, All About Redistricting (Loyola Law School)

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Redistricting

How it is made. A hand-curated reference site on the redistricting process in every state, run by Prof. Justin Levitt with research assistants; it tracks who draws the lines, plan status and litigation. The National Overview offers the full plan table as a CSV. No written method for the table was found beyond the site's description of itself.

Who is in it. All 50 states; congressional, state senate and state house plans for the 2000, 2010 and 2020 cycles (cycle values seen in the CSV and the map download filter, which also lists 2030).

Who is not. Local (county, city) districts; vote or population data.

Revisions. The site is edited as events happen (dated updates); the CSV is re-issued under a new dated file name, so a stored copy must keep its file name.

The provider's method → Its terms →

County Presidential Election Returns 2000-2024

Published by MIT Election Data + Science Lab (university or nonprofit). We download it from Harvard Dataverse (university or nonprofit).

Kind of number

Compilation

How often

Irregular re-releases

License

Public domain (CC0 1.0). Credit: MIT Election Data and Science Lab, County Presidential Election Returns 2000-2024, Harvard Dataverse, doi:10.7910/DVN/VOQCHQ

Status on this site

Research preview

Newest figure we show

Describes elections 2000 to 2016 (the site's later years come from other sources)

Method and terms read

October 15, 2026

Used on

Map and county profiles

How it is made. The Dataverse metadata says only that the dataset contains county-level presidential returns 2000-2024 compiled by MEDSL. The method/codebook and the per-state sources file are in the dataset but sit behind a guestbook, so the method was not read.

Who is in it. County-level presidential returns, 2000 to 2024 (per the description).

Who is not. Not stated in the metadata that could be read.

Revisions. The whole dataset is re-released as numbered versions; version 20.0 (2026-02-25) is current and differs from the version 16 (2025-08-22) the site uses (different UNF fingerprints). What changed is not stated.

The provider's method →

U.S. House returns 1976-2018

Published by MIT Election Data + Science Lab (university or nonprofit).

Kind of number

Compilation

How often

After each election

License

Public domain (CC0 (Dataverse release)). Credit: MIT Election Data and Science Lab, U.S. House 1976-2018

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Redistricting

How it is made. District-level U.S. House returns compiled by MEDSL; the GitHub README names the Clerk of the House as the source. No further method text was found in the documents read.

Who is in it. U.S. House general elections by district, 1976-2024 on Dataverse (1976-2018 on GitHub).

Who is not. Not stated in the documents read.

Revisions. Dataverse issues numbered versions (15 seen, released 2026-03-09); the GitHub copy is older and stops at 2018. We keep the exact GitHub version we used and apply our own documented corrections.

The provider's method → Its terms →

OpenStreetMap

Published by OpenStreetMap contributors (OpenStreetMap Foundation) (university or nonprofit).

Kind of number

Platform or transaction data: map features drawn by volunteers

How often

Continuous; the site uses a dated planet file

License

Allowed with credit (Open Database License (ODbL) 1.0). Credit: © OpenStreetMap contributors

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Data centers

How it is made. A map database built and maintained by volunteer contributors. Contributors are told not to copy from copyrighted sources without permission.

Who is in it. Whatever contributors have mapped; for this project, objects tagged as data centers.

Who is not. Anything unmapped; completeness is not guaranteed.

Revisions. Objects change with every edit; a snapshot must keep its date and each row its OSM id.

The provider's method → Its terms →

Public Utility Data Liberation (PUDL)

Published by Catalyst Cooperative (university or nonprofit).

Kind of number

Compilation: cleaned and joined EIA and FERC filings

How often

Quarterly releases

License

Allowed with credit (CC BY 4.0 (data)). Credit: Catalyst Cooperative, PUDL

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Data centers

How it is made. An open-source ETL pipeline that cleans, links and standardizes public energy data (EIA, FERC, EPA and others) into Parquet, DuckDB and SQLite outputs. Raw inputs are archived on Zenodo with DOIs and each release is tied to specific input versions. Stable releases are versioned by date; nightly builds are less validated.

Who is in it. The data sources listed in the documentation, including EIA Forms 860, 861, 923 and 930, EIA AEO and FERC Form 714.

Who is not. Anything the underlying federal forms do not collect (for example data-center facilities).

Revisions. Each stable release is immutable and versioned; later releases can revise values as the agencies revise their data. Nightly builds change daily and should not be cited.

The provider's method → Its terms →

Presidential results by congressional district

Published by The Downballot (private company).

Kind of number

Compilation

How often

After each election and each new map

License

Allowed with credit (Cite by name with a link). Credit: The Downballot (the-db.co/presbycd)

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

Used on

Redistricting

How it is made. Votes from counties wholly inside one district are assigned to it; in split counties, precinct results are assigned to districts (usually by matching precincts to the concurrent House race), and votes not reported by precinct (often absentee or early) are allocated by stated methods. Only official certified results are used; write-ins are left out except where material. A full state-by-state methodology statement is published.

Who is in it. All 435 districts for each published election and set of lines, from 2008 on.

Who is not. Write-in votes (except in rare cases); some districts where data were unavailable.

Revisions. Calculations are updated when certified results change; the methodology was first published March 2021 and last updated April 2025. Sheets can change, so the package pins each CSV by SHA-256.

The provider's method → Its terms →

FiveThirtyEight election results (House)

Published by FiveThirtyEight (ABC News) (private company).

Kind of number

Compilation: a compilation of state-certified returns

How often

No schedule stated

License

Terms unclear (No license in this repository). Credit: FiveThirtyEight, election-results

Status on this site

Published

Newest figure we show

Election results: each figure describes its election day.

Method and terms read

October 15, 2026

Used on

Redistricting

How it is made. A candidate-level compilation of election results from FiveThirtyEight's database, with a source field per result. The README documents the columns but not how results were collected or checked; the project's own check against the Clerk of the House found and overrode wrong tallies.

Who is in it. Per the README: elections for governor, U.S. Senate, president and (by the office_name values) U.S. House, general and primary, from November 1998. The House file also holds earlier years (the package found 1976-1980 complete).

Who is not. Write-ins are not always tied to a candidate; unopposed races may have no votes.

Revisions. No revision policy stated; races.csv carries created_at and updated_at timestamps, so rows can change. We keep a hashed copy of the file we used.

The provider's method → Its terms →

Indeed Job Postings Index

Published by Indeed Hiring Lab (private company).

Kind of number

Platform or transaction data: job postings on one job site

How often

Daily series, updated weekly

License

Allowed with credit (CC BY 4.0). Credit: Indeed Hiring Lab

Status on this site

Published

Newest figure we show

Describes October 9, 2026

Retrieved

October 15, 2026

Next release

Weekly

Method and terms read

October 15, 2026

Used on

Faster signals

How it is made. Each series is the level of seasonally adjusted Indeed job postings (seven-day trailing average) relative to 1 February 2020 = 100. Each geography is seasonally adjusted separately with a Bundesbank method for daily data, adopted in November 2024, when the history was revised. Postings are Indeed's de-duplicated count of jobs visible on its site, both posted directly and aggregated from other sites.

Who is in it. Job postings shown on Indeed's US site: jobs employers post directly (free or paid) and jobs Indeed aggregates from employer career sites, applicant tracking systems and other websites, after quality review.

Who is not. Openings that are not advertised online, and postings Indeed judges 'low quality'. A posting is not a vacancy count: one posting can cover several openings, a job can stay online after it is filled, and turnover-heavy occupations post more often. No county-level series.

Revisions. The whole history was revised when the seasonal-adjustment method changed in November 2024; seasonal factors for the latest year are projected and can change, and Hiring Lab says methods are re-evaluated periodically (changes are logged on the data portal's 'What's New' page).

The provider's method → Its terms →

VoteHub Polling API

Published by VoteHub (private company).

Kind of number

Compilation: a compilation of published polls

How often

As polls are released

License

Allowed with credit (CC BY 4.0). Credit: VoteHub Polling API, CC BY 4.0

Status on this site

Published

Newest figure we show

Describes polls as of October 7, 2026

Retrieved

October 7, 2026

Method and terms read

October 15, 2026

Used on

Senate polls

How it is made. An aggregation of published polls from many sources served through a free REST API; filters by poll type, pollster, subject, dates, sample size and population. The API documentation describes the data fields but not how polls are selected or checked.

Who is in it. Polls in VoteHub's database, flagged for internal and partisan sponsorship.

Who is not. Inclusion rules are not stated.

Revisions. Not stated.

The provider's method → Its terms →

Zillow Home Value Index (ZHVI)

Published by Zillow Group (private company).

Kind of number

Model estimate: modeled typical home value from Zillow's own home-value estimates

How often

Monthly

License

Terms unclear (Zillow terms of use (aggregate data, with Zillow cited on every page that shows it)). Credit: Data Provided by Zillow Group

Status on this site

Research preview

Newest figure we show

Describes August 2026

Retrieved

September 28, 2026

Method and terms read

October 15, 2026

Used on

Economic score, Map and county profiles

How it is made. ZHVI is built from Zestimates (Zillow's modelled value of each home), not from sales alone. Each month it takes the month-over-month change in Zestimates across the property universe, sets the current level from a trimmed mean of Zestimates (the 35th-65th percentile for the mid-tier series), and chains that level back in time. Since the January 2023 release the full series uses the neural-network Zestimate; before 2012 the new series is stitched to the historical index.

Who is in it. Single-family, condominium and co-operative homes with a county record (the 'all homes' segment), including homes that have not sold recently; 3,071 counties in the county file.

Who is not. Units in 5+ unit rental buildings are outside 'all homes'; homes built after a given month are excluded from that month's universe; homes with extreme Zestimates or data-quality issues may be removed; regions with poor coverage, anomalous changes or under two years of history are suppressed.

Revisions. All series are subject to revision; the whole history was replaced when the neural Zestimate was adopted (January 2023 release).

Required notice. Data Provided by Zillow Group.

The provider's method → Its terms →

US County Level Election Results 08-24 (2024 county presidential results)

Published by Tony McGovern (GitHub: tonmcg) (individual compiler). The underlying data come from Fox News election results pages (scraped) (private company).

Kind of number

Compilation: results scraped from a news site; the compiler calls them not authoritative

How often

After each election

License

Terms unclear (MIT License (written for software)). Credit: tonmcg, US_County_Level_Election_Results_08-24, data scraped from Fox News

Status on this site

Published

Newest figure we show

Describes the November 5, 2024 election

Retrieved

September 28, 2026

Method and terms read

October 15, 2026

Used on

Map and county profiles

How it is made. A compilation scraped from news-site results pages; the 2024 file was scraped from Fox News. The author states the data is not authoritative and asks researchers to verify against primary sources. Alaska is reported by house district and DC by ward, each keyed to a 5-digit FIPS-style code (Alaska districts use a constructed code).

Who is in it. County-level presidential results for all states; Alaska by state house district, DC by ward.

Who is not. Not stated.

Revisions. No revision policy stated.

The provider's method → Its terms →

U.S. Atlas TopoJSON (base map shapes)

Published by Mike Bostock (topojson/us-atlas) (individual compiler). The underlying data come from U.S. Census Bureau cartographic boundary files (2017) (federal agency).

Kind of number

Compilation: Census boundary files converted for web maps

How often

Static (version 3)

License

Allowed with credit (ISC license). Credit: Copyright 2013-2019 Michael Bostock

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

October 15, 2026

How it is made. A TopoJSON conversion of the Census Bureau's 2017 cartographic boundary shapefiles, quantized and simplified; state outlines are made by merging counties and the nation by merging states so the topology is consistent. Projected (Albers USA, 975x610) and unprojected versions are provided.

Who is in it. U.S. counties, states and the nation as in the 2017 Census cartographic boundary files.

Who is not. County changes after 2017 are not in this edition (the README names the 2017 edition only).

Revisions. Earlier editions are in past releases; the site uses version 3.0.1.

The provider's method → Its terms →

Audited 2020 county results and the 2020 descriptive model

Published by Providence Strategy (our research).

Kind of number

Derived by us: official county results checked against state totals, and a fixed model fitted to them

How often

Released once (Outcome Audit v31)

License

Allowed with credit (Our publication). Credit: Providence Strategy

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

Our own method, published on this site

Used on

Map and county profiles

Our method →

County-leverage simulation (Counties in focus)

Published by Providence Strategy (our research).

Kind of number

Model estimate

How often

Released once per model run

License

Allowed with credit (Our publication). Credit: Providence Strategy

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

Our own method, published on this site

Used on

Counties in focus

Our method →

Data-center project register

Published by Providence Strategy (our research).

Kind of number

Compilation: every figure quoted from a downloaded primary document

How often

When the register is extended

License

Allowed with credit (Our publication). Credit: Providence Strategy

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

Our own method, published on this site

Used on

Data centers

Our method →

Historical departures (state results against national swing)

Published by Providence Strategy (our research).

Kind of number

Derived by us

How often

Released once

License

Allowed with credit (Our publication). Credit: Providence Strategy

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

Our own method, published on this site

Used on

Findings

Our method →

County economic score, two lenses

Published by Providence Strategy (our research).

Kind of number

Derived by us

How often

Released once; weights preregistered

License

Allowed with credit (Our publication). Credit: Providence Strategy

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

Our own method, published on this site

Used on

Economic score, Map and county profiles

Our method →

Accepted election release: seats, candidates and district relationships

Published by Providence Strategy (our research).

Kind of number

Compilation: official documents compiled into one checked release

How often

When a release is accepted

License

Allowed with credit (Our publication). Credit: Providence Strategy

Status on this site

Published

Newest figure we show

Election results: each figure describes its election day.

Method and terms read

Our own method, published on this site

Used on

Map and county profiles

Our method →

Faster signals: selection and year-earlier comparisons

Published by Providence Strategy (our research).

Kind of number

Derived by us: the comparisons with a year earlier beside each provider's figure, computed by preregistered rules

How often

With each refresh

License

Allowed with credit (Our publication). Credit: Providence Strategy

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

Our own method, published on this site

Used on

Faster signals

Our method →

Migration corridors by political context

Published by Providence Strategy (our research).

Kind of number

Derived by us

How often

Released once

License

Allowed with credit (Our publication). Credit: Providence Strategy

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

Our own method, published on this site

Used on

Migration

Our method →

Redistricting seats-votes analysis

Published by Providence Strategy (our research).

Kind of number

Derived by us

How often

Released once

License

Allowed with credit (Our publication). Credit: Providence Strategy

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

Our own method, published on this site

Used on

Redistricting

Our method →

Young-adult labor-market context

Published by Providence Strategy (our research).

Kind of number

Derived by us

How often

Released once; rules preregistered

License

Allowed with credit (Our publication). Credit: Providence Strategy

Status on this site

Published

Newest figure we show

Not a recurring series on this site: each figure carries its own date where it is shown.

Method and terms read

Our own method, published on this site

Used on

Young-adult context

Our method →

Regional economic accounts: GDP and personal income by county and state

Published by U.S. Bureau of Economic Analysis (federal agency).

Linked, not shown. Cited only as a reference in the white paper (its interactive tables); no GDP figure is shown. BEA's personal income tables are shown separately, from BEA's public CSV downloads, and checked against BEA's Data API: see the state and county personal income sources and the BEA Data API source.

Kind of number

Model estimate: estimates built from many federal sources

How often

Yearly for counties; quarterly for states

License

Public domain (U.S. government statistics). Credit: U.S. Bureau of Economic Analysis

Status on this site

Linked, not shown

Method and terms read

October 15, 2026

How it is made. County GDP estimates the value of goods and services produced in a county; county personal income counts residents' income from wages, proprietors' income, dividends, interest, rents and government benefits by place of residence. BEA publishes method papers (not read here).

Who is in it. All U.S. counties (about 3,100 in the 2024 release summary).

Who is not. Not stated on the pages read.

Revisions. Each release revises earlier years; the Interactive Data application shows only the latest vintage, so a cited figure should name its release.

The provider's method → Its terms →

Business Trends and Outlook Survey

Published by U.S. Census Bureau (federal agency).

Not used. Read, not yet shown: it is an experimental survey of business conditions by state, and which of its measures belongs beside county figures is an open decision.

Kind of number

Survey sample

How often

Every two weeks

License

Allowed with credit (U.S. government statistics). Credit: U.S. Census Bureau, Business Trends and Outlook Survey

Status on this site

Not used

Method and terms read

October 15, 2026

How it is made. A voluntary, qualitative survey of employer businesses: each year about 1.2 million businesses are sampled (stratified by sector and state) and split into six biweekly panels of about 200,000, each answering about every 12 weeks. Answers are categorical (increased, no change, decreased; excellent to poor), and BTOS publishes the weighted percentage choosing each answer plus 0-100 indexes, with standard errors from a delete-a-group jackknife. Weights are adjusted for nonresponse by state and employment size and by sector and employment size; responses are not edited. Estimates with a relative standard error above 50 percent are suppressed.

Who is in it. Employer businesses (single- and multi-location since September 2023) in the 50 states, DC and Puerto Rico, excluding a list of out-of-scope industries. State figures cover businesses operating in a single state.

Who is not. Non-employer businesses; agriculture production, railroads, the Postal Service, public administration, private households, religious and civic organizations and some financial vehicles. Businesses operating in more than one state are excluded from every state total (shown as 'XX'). No county geography; MSA figures only for the 25 most populous MSAs. Before September 2023 only single-location businesses were covered, so earlier ('V1') files are not comparable.

Revisions. The JavaScript-rendered About page says 'The BTOS does not do revisions.' and that edits and benchmarking are not performed. That page cannot be quote-checked (its fetched text is an empty JavaScript shell), so the only checkable support is that responses are not edited and that late responses are not tabulated. A new sample is drawn every year, and the questionnaire changed over time (for example the AI questions in November 2025), so long comparisons need care.

The provider's method → Its terms →

Chicago Fed Labor Market Indicators

Published by Federal Reserve Bank of Chicago (federal reserve system).

Linked, not shown. National only, with no data file found, and republication needs written permission: linked as reading.

Kind of number

Model estimate

How often

Twice a month

License

Terms unclear (Site terms). Credit: Federal Reserve Bank of Chicago

Status on this site

Linked, not shown

Method and terms read

October 15, 2026

How it is made. Flow rates into and out of unemployment from CPS data, scaled to match the BLS unemployment rate, combined with real-time indicators (UI claims, Morning Consult, JOLTS, Conference Board, Indeed and Lightcast postings, ADP and others) in a partial least squares model to forecast the BLS unemployment rate.

Who is in it. U.S. labor market (national).

Who is not. No state or county detail.

Revisions. Not stated on the pages read.

The provider's method → Its terms →

Household Debt and Credit

Published by Federal Reserve Bank of New York (federal reserve system). The underlying data come from Equifax credit reports (New York Fed Consumer Credit Panel) (private company).

Not used. Read, not yet shown: it is published by state once a year (latest: end of 2025), not a fast signal, and whether Equifax's ownership limits republishing is not stated.

Kind of number

Administrative records: a 5% sample of anonymized Equifax credit reports

How often

Quarterly report; all states once a year

License

Allowed with credit (New York Fed terms of use). Credit: New York Fed Consumer Credit Panel / Equifax

Status on this site

Not used

Method and terms read

October 15, 2026

How it is made. Aggregates from the New York Fed Consumer Credit Panel, an anonymized, nationally representative 5% random sample of individuals with a Social Security number and an Equifax credit report, plus everyone at the same addresses; most report figures use a 0.1% subsample. Per-capita figures divide totals by the number of people in the sample, so they apply to adults with a credit report. Figures are nominal.

Who is in it. Individuals with a Social Security number and an Equifax credit report (usually aged 19 and over), and others living at the same address. All 50 states plus Puerto Rico in the annual state file; 11 selected states in the quarterly report.

Who is not. Adults without a credit file or without a Social Security number; accounts not reported by creditors in the last 3 months; authorized-user, disputed, medical and commercial trades. No data below the state level are published, and quarterly data for all states are not published, both because of contractual restrictions with the data provider.

Revisions. The documents read describe no routine revision schedule. They do record breaks: student-loan delinquency jumped in 2012Q3 when defaulted loans reappeared, collection accounts fell after a 2017Q4 reporting change, the credit score changed to VantageScore 4.0 from 2026Q1, and two state transition-rate pages carry 'Revised May 2017'. The state file warns its totals may not match the quarterly report because of sampling variation.

The provider's method → Its terms →

AAA Gas Prices, state averages

Published by AAA (American Automobile Association) (university or nonprofit). The underlying data come from OPIS, with WEX, Inc. (station survey) (private company).

Linked, not shown. AAA's terms allow personal, non-commercial use only and forbid reproducing or displaying its content. Its state averages are withdrawn from the site's cost tiles and linked instead. Where the U.S. Energy Information Administration publishes a state or its region, the weekly EIA price is shown instead, labeled State or Region.

Kind of number

Platform or transaction data: prices reported from a survey of stations and payment-card transactions

How often

Daily

License

Not allowed without permission (AAA terms of use). Credit: AAA

Status on this site

Linked, not shown

Method and terms read

October 15, 2026

Used on

Economic score, Map and county profiles

How it is made. AAA's site is updated daily by OPIS with average retail prices from a survey of up to 120,000 stations run with WEX, Inc. No further method (weighting, station sample) was found on the pages read.

Who is in it. Retail stations in the OPIS/WEX survey; 50 states and DC.

Who is not. Not stated.

Revisions. The page shows only the current day; no history or revision policy is published.

The provider's method → Its terms →

Ballotpedia

Published by Lucy Burns Institute (Ballotpedia) (university or nonprofit).

Linked, not shown. Linked as corroboration only; its terms could not be read from our environment.

Kind of number

Compilation

How often

Continuous

License

Terms unclear (Not read (bot challenge)). Credit: Ballotpedia

Status on this site

Linked, not shown

Method and terms read

Not readable from our environment now; our data package recorded the documentation it used on October 15, 2026

Used on

Elections

How it is made. Not read: every ballotpedia.org page answered with a bot challenge.

Who is in it. Not read.

Who is not. Not read.

Revisions. Not read.

The provider's method →

ADP National Employment Report

Published by ADP Research (private company). We download it from FRED, Federal Reserve Bank of St. Louis (federal reserve system).

Linked, not shown. ADP's own terms forbid copying or publishing its information for public or commercial purposes without written consent, and FRED's terms defer to the owner's.

Kind of number

Platform or transaction data: payroll records of ADP's clients, re-weighted each year to federal benchmarks

How often

Monthly

License

Not allowed without permission (ADP terms of use). Credit: ADP National Employment Report

Status on this site

Linked, not shown

Method and terms read

October 15, 2026

How it is made. Weekly snapshots of ADP payroll data for a matched sample of business establishments (ADP payroll accounts) give week-to-week employment growth, which is weighted to the QCEW distribution of employment by industry x state x establishment size, turned into an index, and applied to a QCEW base level. ADP describes it as an independent measure, not a forecast of the BLS payroll report. The sample is ADP's clients: more than half a million companies with more than 26 million employees.

Who is in it. Private-sector employees on the payrolls of ADP client companies (more than 26 million employees), reweighted to QCEW.

Who is not. Government employment; employers that do not use ADP payroll (represented only through the QCEW weighting); clients reporting no activity in a month are excluded from that month's estimate. No geography below census division is published.

Revisions. The prior month is revised when late client data arrive, and every February the series is reweighted to QCEW benchmark data through March of the previous year.

The provider's method → Its terms →

Apartment List rent estimates

Published by Apartment List (private company).

Not used. Public CSV downloads exist, but the site terms allow personal, non-commercial use only; not used unless Apartment List confirms otherwise.

Kind of number

Model estimate

How often

Monthly

License

Terms unclear (Site terms). Credit: Apartment List

Status on this site

Not used

Method and terms read

October 15, 2026

How it is made. ACS median rents for recent movers are moved forward to the current month with a same-unit repeat-transaction growth rate from leases on Apartment List's platform; the vacancy index uses only properties that list on the platform.

Who is in it. Locations meeting sample-size rules (national, states, hundreds of metros, counties and cities).

Who is not. Locations below the sample threshold.

Revisions. Not stated on the pages read.

The provider's method → Its terms →

Bank of America Institute reports

Published by Bank of America Institute (private company).

Not used. PDF reports with all rights reserved; no downloadable series.

Kind of number

Platform or transaction data

How often

Monthly reports

License

Terms unclear (All rights reserved). Credit: Bank of America Institute

Status on this site

Not used

Method and terms read

October 15, 2026

How it is made. Reports built on Bank of America's internal data; no methodology was found on the pages read.

Who is in it. Bank of America customers.

Who is not. Non-customers.

Revisions. Not stated.

The provider's method →

Homebase small business data

Published by Homebase (private company).

Not used. Published as PDF reports, and its terms grant no right of reuse.

Kind of number

Platform or transaction data

How often

Monthly

License

Terms unclear (Site terms). Credit: Homebase

Status on this site

Not used

Method and terms read

October 15, 2026

How it is made. Monthly tallies from Homebase customers' timesheets (employees working, hours, wages, businesses open, hires and departures), compared with the same window a year earlier; not seasonally adjusted.

Who is in it. Small businesses using Homebase.

Who is not. Everyone else.

Revisions. Not stated.

The provider's method → Its terms →

JPMorganChase Institute research

Published by JPMorganChase Institute (private company).

Linked, not shown. Research reports only, with no downloadable series, and its terms bar reuse without written consent: linked as reading.

Kind of number

Platform or transaction data: research drawn from the bank's own account data

How often

Irregular reports

License

Not allowed without permission (Site terms). Credit: JPMorganChase Institute

Status on this site

Linked, not shown

Method and terms read

October 15, 2026

How it is made. Analyses built on JPMorganChase's proprietary data; methods are described per report (none read here).

Who is in it. JPMorganChase customers (per report).

Who is not. Non-customers.

Revisions. Not applicable to reports.

The provider's method → Its terms →

Lightcast

Published by Lightcast (private company).

Not used. Not evaluated: the project's brief lists it among providers to skip (paid, login or bot-protected data). The project uses only series it can download and whose terms it has read.

Kind of number

Platform or transaction data

How often

Not evaluated

License

Terms unclear (Not evaluated). Credit: Lightcast

Status on this site

Not used

Method and terms read

Not read

LinkUp

Published by LinkUp (private company).

Not used. Not evaluated: the project's brief lists it among providers to skip (paid, login or bot-protected data). The project uses only series it can download and whose terms it has read.

Kind of number

Platform or transaction data

How often

Not evaluated

License

Terms unclear (Not evaluated). Credit: LinkUp

Status on this site

Not used

Method and terms read

Not read

The New York Times poll feed

Published by The New York Times (private company).

Linked, not shown. Never shown as figures. It was read on October 7, 2026 to cross-check our polls and label candidates' parties; since October 15 it is no longer fetched, because its terms allow personal, non-commercial use and ask for written consent before automated collection. The facts derived on October 7 are kept.

Kind of number

Compilation

How often

As polls are released

License

Not allowed without permission (New York Times terms of service). Credit: The New York Times

Status on this site

Linked, not shown

Method and terms read

October 15, 2026

Used on

Senate polls

How it is made. No method document for the feed was found. The file is one row per poll question and candidate answer, with pollster, dates, population, party and percentage.

Who is in it. U.S. Senate polls (per file name and rows).

Who is not. Not stated.

Revisions. Not stated.

The provider's method → Its terms →

Paychex Small Business Employment Watch

Published by Paychex, Inc. (private company).

Not used. Charts only, with no downloadable series and no terms of use found.

Kind of number

Platform or transaction data

How often

Monthly

License

Terms unclear (No terms found). Credit: Paychex

Status on this site

Not used

Method and terms read

October 15, 2026

How it is made. A same-store index of year-over-year change in worker counts at small Paychex clients (fewer than 50 workers), from a five-week payroll window each month, after outlier removal; 100 means no year-over-year change. Reported for the U.S., regions, the 20 largest states and metros and eight NAICS supersectors.

Who is in it. Paychex clients with fewer than 50 employees (about 350,000).

Who is not. Non-Paychex employers and firms with 50 or more workers; states and metros outside the 20 largest.

Revisions. No monthly revisions; periodic restatements possible.

The provider's method → Its terms →

Placer.ai

Published by Placer.ai (private company).

Not used. Not evaluated: the project's brief lists it among providers to skip (paid, login or bot-protected data). The project uses only series it can download and whose terms it has read.

Kind of number

Platform or transaction data

How often

Not evaluated

License

Terms unclear (Not evaluated). Credit: Placer.ai

Status on this site

Not used

Method and terms read

Not read

Realtor.com residential listings data

Published by Realtor.com (Move, Inc.) (private company).

Linked, not shown. The data page asks for credit, but the terms of use forbid republishing without written permission, and the terms say that rule always applies.

Kind of number

Platform or transaction data

How often

Monthly

License

Terms unclear (Move, Inc. terms of use). Credit: Realtor.com Economic Research

Status on this site

Linked, not shown

Method and terms read

October 15, 2026

How it is made. Counts and medians computed from MLS listings of existing single-family homes, condos, townhomes, row homes and co-ops shown on Realtor.com. Active listing count is the typical daily number of active (not pending) listings in the month; new listing count is listings added in the month; median days on market runs from first listing to closing or removal. These are asking-price and listing measures, not sales.

Who is in it. MLS-listed for-sale existing homes (single-family and condo/townhome/row home/co-op) that reach Realtor.com; 3,119 counties in September 2026.

Who is not. New construction not listed through a contributing MLS; for-sale-by-owner homes not on an MLS. Figures are least reliable in small geographies and in markets with partial listing coverage or changing definitions of active inventory; 879 of 3,119 counties carried quality_flag = 1 in September 2026.

Revisions. The full history is reissued every month and past values may change; data were restated in January 2025, so files downloaded before then are not comparable. Each release must be kept and the whole series replaced at every refresh.

The provider's method → Its terms →

Redfin Data Center

Published by Redfin (private company).

Linked, not shown. No page grants permission to republish its figures, and the terms of use reserve all data to Redfin.

Kind of number

Platform or transaction data

How often

Monthly for counties; weekly for metros

License

Terms unclear (Redfin terms of use). Credit: Redfin

Status on this site

Linked, not shown

Method and terms read

October 15, 2026

How it is made. Counts and medians of residential listings and closed sales by period. Homes sold counts closings by close date; pending sales count homes going under contract; median sale price is the median final sale price of homes closed in the period; active listings count homes for sale at any point in the period and inventory counts homes for sale on its last day (both exclude listings over a year old). The most recent period is pre-adjusted for sales Redfin expects to be reported late.

Who is in it. Residential properties (single family, condo/co-op, townhouse, multi-family 2-4 units) in Redfin's data; 3,097 counties in September 2026 (2,902 with sales figures).

Who is not. Counties with no recorded activity in a month (values 'NA'); 5+ unit buildings. Counties are not offered at weekly frequency.

Revisions. All data are subject to revision: history is revised whenever late records arrive (mostly within a few months), and the latest month is pre-adjusted for expected late reports for about four weeks. Seasonal factors are re-estimated at every update, so adjusted history can shift.

The provider's method → Its terms →

Revelio Labs public labor statistics

Published by Revelio Labs (private company).

Not used. Called freely available, but its terms limit use to personal or internal business purposes and bar copying.

Kind of number

Model estimate

How often

Monthly

License

Not allowed without permission (Site terms). Credit: Revelio Labs

Status on this site

Not used

Method and terms read

October 15, 2026

How it is made. Employment, hiring, attrition, postings and posted salaries estimated from deduplicated professional profiles, corrected for reporting lags and reweighted to BLS OEWS occupation, location and industry benchmarks; seasonally adjusted with STL.

Who is in it. Workers with observable professional profiles, reweighted to the U.S. workforce.

Who is not. Military, interns, freelancers, self-employed, full-time students and others the method excludes.

Revisions. History is revised as new profile data arrive.

The provider's method → Its terms →

Truth Social posts

Published by Trump Media & Technology Group (private company).

Linked, not shown. Cited by link as evidence of an endorsement statement; its terms could not be read from our environment.

Kind of number

Platform or transaction data

How often

As posted

License

Terms unclear (Not read (bot challenge)). Credit: Truth Social, the post's author

Status on this site

Linked, not shown

Method and terms read

Not readable from our environment now; our data package recorded the documentation it used on October 15, 2026

Used on

Elections

How it is made. Not applicable; individual posts are cited as evidence of endorsements.

Who is in it. Not applicable.

Who is not. Not applicable.

Revisions. Not applicable.

The provider's method →

Zillow Observed Rent Index (ZORI) and listings data

Published by Zillow Group (private company).

Not used. Its terms do not clearly allow republishing the research downloads, so new Zillow series are not added until Zillow or the owner settles it.

Kind of number

Platform or transaction data

How often

Monthly

License

Terms unclear (Zillow terms of use; no license on the research downloads). Credit: Data Provided by Zillow Group

Status on this site

Not used

Method and terms read

October 15, 2026

How it is made. ZORI is a repeat-rent index: it measures the change in asking rent for the same rental unit listed more than once on Zillow, with listing pairs reweighted so that building age and size shares match the American Community Survey rental stock. The index is smoothed with a three-month moving average and given a dollar level equal to the weighted mean of asking rents between the 35th and 65th percentile in the latest month, chained back in time. It measures asking (listed) rents, not rents paid by sitting tenants.

Who is in it. Rental units listed for rent on Zillow more than once (repeat-rent pairs), all homes plus multifamily (5+ unit buildings) in the file used; published only where Zillow judges data sufficient. 1,407 counties in the county file (1,406 with an August 2026 value).

Who is not. Counties with too few repeat rental listings (roughly half of U.S. counties have no county row); rents of sitting tenants and units never listed on Zillow, except through the ACS reweighting. Series that fail Zillow's quality checks are not published.

Revisions. The method page states no revision policy. It says the dollar level is set from the latest month and chained back through the index's month-to-month changes, so earlier dollar values are tied to the newest release; whether past values change from one release to the next is not stated (keep each release's file).

The provider's method → Its terms →