The method behind the numbers
Our standard is simple: you should be able to reproduce a number and understand its limits.
Source selection
We prioritize primary government datasets, followed by primary academic research. Every observation records its source, unit, period, population, geography, revision, and release. Research and interpretation are kept separate from the statistical record.
Normalization & calculations
We preserve source precision. Counts published in thousands are multiplied by 1,000; rounded values are never presented as unrounded originals. Native-born share is calculated as 100 minus foreign-born share. Published percentages take precedence over ratios of rounded counts. Percentage-point change and percentage change are labeled separately.
Population denominators
A foreign-born share uses all residents as its denominator. Naturalized share in this release uses foreign-born residents. The two percentages are not interchangeable. A rate requires a numerator and denominator referring to comparable populations, geography, and time.
Estimates & uncertainty
ACS values are sample estimates. Table 1 margins of error describe 90% confidence intervals. We display available margins of error and do not manufacture uncertainty for observations that omit it. Differences between estimates require statistical testing before claims of significance.
Missing values
Unknown is not zero. Missing years, unsupported geographic detail, and unimported metrics are shown as unavailable. Historical charts plot observed years; we do not interpolate missing values. The latest year in this release is distinct from the latest release an agency may have published.
Historical comparability
Decennial Census and ACS estimates use different collection methods. Census changed classification of American Samoa nativity before the 2018 ACS. Table 1 and Table 2 of ACSBR-019 exclude people born at sea; historical Figure 1 has a different scope. We retain these qualifications with the data.
Revisions & publication
Original files are preserved under an agency/dataset/retrieval-date path and identified by a SHA-256 hash. New observations are staged and validated. Revisions append a new version without deleting the previous one. Only an authorized administrator can publish a validated run; anomalies require a review explanation.
H-1B counting
USCIS petitions, DOL labor condition applications, Department of State visas, and individual workers count different events or people. A person can have multiple petitions. Filings and approvals during one fiscal year are not the same cohort, so approvals divided by filings is not an approval rate. USCIS reported compensation does not measure realized earnings.
Border statistics
Encounters can include repeated events involving one person. USBP apprehensions, OFO inadmissibles, and historical Title 42 expulsions must retain their categories. Monthly calendar dates and federal fiscal years are stored explicitly. Encounters do not estimate the resident unauthorized population.
Crime and fraud safeguards
Citizenship, birthplace, ethnicity, nativity, and immigration status are different variables. Federal statistics do not represent all U.S. crime, and state findings are not extrapolated nationally. Comparable crime rates require comparable denominators. Allegations, investigations, charges, and convictions remain separate. Asylum denial is never labeled fraud.
Review status & scope
Reviewed releases include 2023–2024 ACS nativity, 2024 country-of-birth detail where published, and FY2025 DOL employer filings. Historical and topic gaps remain documented on the Coverage page. Unavailable data are never replaced with demo statistics.