Sources / methods / limitations
Open data.
Visible methods.
Every number should have a definition, a date and a way to check it.
What the dashboard measures
US Data Watch tracks U.S. national estimates of everyday prices, consumer prices, earnings and unemployment. Prices cover defined products and units. White bread is priced per pound, milk per gallon and eggs per dozen. Electricity is a unit price, not a household bill.
The rent series is a price index. It is not an average dollar rent or the price of a new apartment listing. Hourly earnings are an average, not a median, and are shown in nominal dollars.
Our sources
The dashboard uses 36 source series and one derived net payroll-change series from the Bureau of Labor Statistics Public Data API. Every metric links to its original series. History starts in 2008 where observations exist. BLS average prices, CPI indexes and commodity producer price indexes are not seasonally adjusted. Employment, labor turnover and earnings series are seasonally adjusted. Median weekly earnings are quarterly averages; other displayed series are monthly. Source counts published in thousands are converted to jobs, people or events, with units recorded. We preserve published footnotes and unavailable values.
BLS average-price methodology explains how product mix and sample changes can affect price-level comparisons. Changes in average prices are not interchangeable with a CPI inflation measure.
Updates and revisions
The data cache is checked once daily. Source series publish monthly or quarterly; a daily check does not create a new observation. “Observed” describes the period the number measures. “Checked” describes when we fetched the source. Failed updates retain the last valid snapshot. Revisions are reflected in the latest data; dated snapshots are retained.
Charts break at missing reporting periods. Quarterly observations are shown as quarters, without monthly interpolation. Missing values are never zero-filled or interpolated. A metric can have an older latest observation than another metric. Use its source details to see the date and footnotes.
Administration comparisons
Each presidential term starts in January. The default baseline is the last full month before inauguration: December. This avoids treating a mixed-administration January average as a clean starting point. January observations still include both administrations; this limitation cannot be removed from monthly data.
The card defaults compare each term’s latest available observation with its baseline, the same month one year earlier, or the previous month. Historical term defaults end in December of the final full year. Card month selectors can override both observation dates. Choosing None removes the computed comparison. Each economic card opens its selected monthly observations, with an additional view to compare two terms. By default the endpoints use the same number of elapsed months. Turning off “Match elapsed months” allows independent dates, which can span different lengths of time. All changes require exact observations at both ends.
Rate change = latest rate − baseline rate, in percentage points
Net payroll change = current employment − exact previous-month employment
Net-change comparison = endpoint monthly change − baseline monthly change, in jobs
Comparisons are descriptive. They do not demonstrate that a president or administration caused an outcome. No composite administration score is calculated.
Non-food commodities
The Commodities tab tracks U.S. domestic producer selling-price changes using BLS commodity Producer Price Indexes. Each index uses its verified base period, shown beside the value. These are indexes, not dollars per pound, board, ton or sheet. They do not describe retail checkout prices, futures or spot quotes, and domestic output coverage excludes imports.
Compare percentage changes within a material over time. Comparing index levels across different materials does not reveal which material costs more. Copper and brass mill shapes include brass; gypsum covers more than drywall; cement and ready-mix concrete are distinct measures. Lower values are shown as lower input-price pressure, not as better outcomes for every producer. Recent releases are preliminary and may be revised. BLS PPI methodology.
Jobs and income
Payroll jobs, employed people, hiring flows and vacancies measure different things. Net payroll change is calculated from adjacent BLS employment levels; positive values mean net additions, and negative values mean net losses. Hires and layoffs are separate JOLTS flows. Subtracting layoffs from hires does not reproduce net payroll change, and farm-related household employment must not be added to nonfarm payroll counts.
Farm-related employment covers agriculture and related industries, including forestry, fishing and hunting; it is broader than farm-only hired workers. Average pay and median pay have different worker coverage. Real earnings use constant 1982-1984 dollars and CPI-U to measure purchasing power. Median usual weekly earnings cover full-time wage and salary workers, excluding the self-employed. They are quarterly averages, not household income or annualized earnings. The baseline before a presidential term is Q4 of the preceding year. Q1 includes days before inauguration, and unreported quarters remain unavailable.
Immigration reporting
The immigration tab uses free official CBP, ICE, State Department, DHS and Justice Department datasets and reports. Every observation carries its reporting window and original source. Encounters count recorded events, not all illegal entries or unique people. ICE completed removals differ from court removal orders. H-1B petition approvals differ from visa issuances and entries; renewals and changes of employer can produce multiple petitions for one worker.
Fiscal years run October 1 to September 30, so a fiscal year can span administrations. Partial fiscal-year counts are compared only with the same reporting cutoff, never with a full-year count. Snapshot stocks, such as pending court cases, are distinct from flows over a period. Agency estimates of self-deportation are labeled and are excluded from computed comparisons and combined totals. Voluntary-departure court decisions do not establish that the departure occurred. Report figures remain available until a newer official release is verified.
Health reporting
Health uses free CDC/NCHS APIs and original AHRQ MEPS-IC reports. Insurance premiums are annual private-sector employer plan averages, with employee contributions and deductibles kept separate. They are not Marketplace quotes. NHIS provides self-reported prevalence and access measures; NHANES examination estimates retain their multi-year cycles, age groups and age adjustment. The uninsured measure covers ages 18 to 64 under its source footnote.
Final birth and death statistics remain separate from provisional estimates. Birth rates use total population; general fertility uses females ages 15 to 44. Age-adjusted disease death rates measure mortality, not prevalence or new cases. Infant mortality uses live births as its denominator. Missing years remain missing. Source failures retain observations and successful check times. Rates alone do not establish statistical significance, individual risk or administration causation. Health information is not medical advice.
Coverage still in progress
A dollar rent estimate can be added from the Census American Community Survey with a free API key. It is annual median gross rent and includes renter-paid utilities, so it will be kept separate from the monthly rent index. The Health tab includes verified employer insurance premiums and separate employee contributions, with annual reporting dates.
Corrections
Use “Report a data issue” in a metric’s source panel. Include the observation date and original source when possible. A submitted report is recorded for review; it does not automatically change published data.