Civilian noninstitutionalized population age 25–64
Bachelor's degree or higher, High school or higher, Some college or higher, Associate degree or higher, Master's degree or higher, Doctoral or professional degree, All education attainment levels
Methods and limitations
How to interpret lab data
The lab uses American Community Survey data to estimate outcomes for groups of people. This page explains who is included, how the estimates are calculated, and when results are too uncertain to report.
Source data
Results come from the U.S. Census Bureau’s American Community Survey Public Use Microdata Sample (ACS PUMS). These records represent a sample of people surveyed across the country. The lab uses them to calculate estimates for groups of people.
The Census Bureau makes PUMS files and their documentation publicly available. The lab does not serve individual survey records. Results include source links and version information so you can trace the data and methods used to calculate them.
Population and terminology
The ACS asks whether a person is deaf or has serious difficulty hearing. In these data, people with a “yes” response are grouped as deaf, and those with a “no” response as hearing. NDC uses deaf as an inclusive term for people who may identify as Deaf, deafblind, deafdisabled, hard of hearing, late-deafened, or hearing impaired.
The ACS does not ask about Deaf cultural identity or sign-language use. This question cannot capture the full range of people's identities or experiences. Read the estimates alongside community knowledge and local research. See NDC's explanation of the term deaf.
Each measure covers a specific population, called its universe. The groups below show who is included. Open a measure's definition for its age range and other inclusion rules.
Civilian noninstitutionalized population age 16–64
Postsecondary enrollment, Employment rate, Employment rate trend, Employment by educational attainment, Unemployed population share, Not in the labor force, All employment statuses
Civilian noninstitutionalized bachelor's degree holders age 25–64
STEM field of degree
Employed civilian noninstitutionalized population age 16–64
Private-sector workers, Private for-profit workers, Private nonprofit workers, Local government workers, State government workers, Federal government workers, Self-employed, business and unpaid family workers, Self-employment, Business ownership, All worker classes
Employed civilian noninstitutionalized population age 16–64 with valid weeks and usual hours worked
Full-time, year-round employment
Full-time, year-round civilian noninstitutionalized workers age 16–64 with valid adjusted earnings, including zero and net losses
Median annual earnings, Median earnings by educational attainment, Mean annual earnings, Mean earnings by educational attainment
Employed noninstitutionalized population age 16–64, including Armed Forces, with a reported occupation
Occupation distribution, Bachelor’s degree or higher by occupation
Civilian noninstitutionalized population age 18 and older
Median age, Population by race and ethnicity, Population by sex reported in the ACS
Deaf civilian noninstitutionalized population age 18 and older
Additional disability
Civilian noninstitutionalized employed population age 16–64 with a valid occupation
STEM occupation employment share
Civilian noninstitutionalized population age 25–64 with a bachelor’s degree or higher and a valid SCIENGP flag
Science and engineering degree share
Civilian noninstitutionalized bachelor’s degree holders age 25–64 with a mapped first field of degree
College graduates by major, Employment by college major
Full-time, year-round employed noninstitutionalized population age 16–64, including Armed Forces, with valid adjusted earnings, including zero and net losses and a reported occupation
Median earnings by occupation
Survey periods and geography
A 1-year estimate uses responses collected during one calendar year. A 5-year estimate combines responses collected over five years; it does not describe the final year alone. The lab uses the following ACS datasets for each geography. Available years and measures depend on the data release.
| Geography | ACS dataset | Period label | Important limitation |
|---|---|---|---|
| United States | ACS 1-year | Single year, such as 2024 | 2020 is omitted because Census did not release standard ACS 1-year estimates. |
| State and District of Columbia | ACS 5-year | Full span, such as 2020–2024 | Consecutive five-year periods share four years of survey responses. |
| Public Use Microdata Area (PUMA) | ACS 1-year or 5-year | Documented annual or full five-year span | Each release identifies its native 2010 or 2020 PUMA boundaries. Matching numeric codes across vintages do not establish equal boundaries. Historical comparisons require matching Census-declared native definitions, audited reference tract identities, and nonoverlapping periods. |
This version does not provide county, city, census tract, ZIP code, metropolitan area, or Puerto Rico estimates. PUMA estimates should not be treated as county estimates. Keep 1-year and 5-year results separate when making comparisons.
Weights and uncertainty
Person weights
Each survey record has a person weight, PWGTP, that represents how many people
in the population it stands for. Estimates use these weights within the population
defined for each measure.
Replicate weights
All 80 replicate
weights, PWGTP1 through PWGTP80,
are used to estimate sampling uncertainty. Each reported estimate and comparison between
populations in the same survey period is recalculated with every replicate weight. The
variation across these calculations is used to determine the standard error.
Reported results include a margin of error and confidence interval at the 95% confidence level. A wider interval means more sampling uncertainty. Result details also include the standard error, number of survey records, and weighted denominator where applicable. Margins of error describe sampling uncertainty; they do not capture every possible source of error, such as inaccurate responses or people not responding to the survey.
Earnings are adjusted with ADJINC and
reported in dollars for the final year of the ACS period. Historical changes convert both
endpoints to a common dollar year using documented BLS R-CPI-U-RS inflation factors. Downloads identify the dollar year, factors, and their version. Unsupported year
selections are explicitly not comparable. Annual ACS earnings refer to the preceding 12
months, so adjacent releases have overlapping income reference months and do not represent
exact calendar-year economic conditions.
Suppression and reliability
351
Minimum sample size
A result is withheld when fewer than 351 survey records are included in its denominator.
30%
Uncertainty warning
Relative standard error (RSE) is the standard error divided by the estimate's absolute value. A source estimate with an RSE above 30% is marked unstable. Historical changes inherit their input estimates' reporting screens; the interval for the change describes uncertainty about its direction.
95%
Confidence level
The lab uses a 95% confidence level for its margins of error and confidence intervals.
The sample-size and RSE thresholds are dashboard reporting rules. “Stable” means an estimate passes the current screen; it does not establish absence of bias, that intervals attain their stated confidence level, or scientific approval. Inference for medians with signed replicate weights or many repeated values remains under review.
When an estimate is withheld, any comparison that needs it is also unavailable. A dash means no value is reported. Check the accompanying status and reason to understand why.
A weighted median is also withheld when the replicate method produces a standard error at
or very near zero. This can happen when the median changes little across replicate
weights, and it does not establish that the estimate has no uncertainty. The result
carries the reason code median_zero_replicate_variance.
Counts, rates and means with a zero replicate standard error are likewise withheld, with
the reason code zero_replicate_variance. Releases built under the earlier ndc-observatory-2026.2 methodology predate this rule and are kept unchanged for
reproducibility: in them, such an estimate (for example a rate of exactly 0%) can appear as
“Stable” with a margin of error of 0. Treat a zero margin of error as missing uncertainty, not
as certainty.
Comparisons and changes
For deaf and hearing comparisons, differences subtract the hearing estimate from the deaf estimate. Ratios divide the deaf estimate by the hearing estimate. A positive difference means the deaf estimate is higher; a negative difference means it is lower.
| Calculation | Interpretation |
|---|---|
| Difference | One estimate minus another, in dollars, people, or years. Differences between rates are expressed in percentage points. |
| Percentage-point difference | One percentage minus another. For example, 60% minus 50% is 10 percentage points. |
| Ratio | One estimate divided by another. A ratio of 1 means the estimates are equal. The result is unavailable if the denominator is zero. |
| Percent change | The change divided by the earlier estimate, multiplied by 100. For example, a rise from 50% to 60% is a 20% increase. |
Check that the measure, ACS dataset, geography boundaries, population, and definition version are compatible before comparing results. The explorer limits the available comparisons to those supported by the selected release.
Trend charts show each period's estimates and any available deaf–hearing comparisons for that period. The historical changes page calculates supported changes between nonoverlapping periods from the same ACS product. Approximate 95% confidence intervals describe one comparison at a time; they are exploratory and do not control error across many comparisons. Results retain input suppression and reliability warnings. Overlapping periods are explicitly not comparable because their cross-period covariance cannot be reconstructed from the available public replicate files. See the Census Bureau's guidance on comparing ACS data.
Methodology version
This page describes ndc-observatory-2026.3 (Deaf Data Observatory methodology (2026.3)). Each measure's
definition specifies its age range and any exclusions, such as Armed Forces members or people
living in institutional group quarters. Differences in these rules, occupation categories, or
earnings calculations can produce estimates that differ from the previous NDC dashboard.
Each data release records the methodology and definition versions used to calculate its estimates. Corrections receive a new version and notes explaining what changed. Include the version and survey period when citing a result so others can identify the same estimates.
Classification clarification for definition versions 2026.8 and later: college-major measures assign the first reported bachelor’s field to one of 14 Census groups. An older five-group precedence description retained in the frozen methodology does not describe these measures. Changes to Census field codes in 2021 and 2023 make comparisons across those dates unavailable for the affected groups; other groups remain eligible when all comparison rules are satisfied.
Browse measure definitions