How your estimate is built.
Start simple. Open every source.

Who the calculator counts

The calculator estimates how many adults in your selected age, sex, and location group match your filters. It does not count active daters or compatible partners.

It starts with the 2024 American Community Survey (ACS), a Census survey of people ages 20–69. National and state results use Census weights. The technical dataset name is the one-year Public Use Microdata Sample (PUMS).

One limit is that the ACS includes people in homes and group living settings. The CDC health data covers civilians who do not live in institutions. The sources therefore cover slightly different groups. A future rebuild will align or separately classify them.

The calculator does not claim that every person counted is single, actively dating, attracted to the user, geographically reachable, or mutually compatible. The optional relationship filter means not currently married; it does not prove active availability.

What Census records measure together

Census records measure age, annual personal income, education, sex, state, and current marital status for the same person. The calculator counts these filters together. It does not multiply separate headline percentages.

The source uses the ACS binary SEX variable, with codes for male and female. This interface labels those groups as men and women. The source does not measure gender identity, sexual orientation, attraction, or intended partner eligibility.

Before income thresholds are applied, PINCP is converted to the file's constant-dollar basis using the record's ADJINC factor. Education thresholds follow the ACS SCHL codes. High school begins with a regular diploma or equivalent. Some college includes any postsecondary attendance. Bachelor’s includes bachelor’s or higher. Graduate includes master’s, professional, or doctoral degrees.

The generated data tables group records into age, income, education, sex, and marital-status cells. The public generator is versioned with the site so the model can be rebuilt from the official source ZIP.

Height is modeled from measurements

The calculator uses a health survey to estimate what share of people meet a selected height minimum. It must combine that estimate with Census data because the Census does not measure height.

Height uses measured standing-height records from the August 2021–August 2023 NHANES examination files. Model version 2026.07-evidence-v3 groups adults by sex and decade of age, applies the CDC MEC examination weight, and calculates each group’s weighted mean and population-distribution standard deviation.

The calculator fits a normal height distribution to each weighted group and uses its surviving share at offered cutoffs from 4′8″ through 6′6″. The checked-in model includes record counts and effective sample sizes, and its tests compare the fitted curves with all published CDC 5th–95th-percentile anchors. When a modeled tail represents fewer than five effective observations, v3 suppresses the exact point and reports a one-sided rarity floor — “1 in X or rarer,” at most X% of the age group — set by an approximate 95% Wilson upper bound based on effective sample size.

That upper bound is a stability guard, not a full survey confidence interval: it does not account for every NHANES design feature or uncertainty in the fitted normal shape. Because Census records do not include measured height, the height rate is also applied independently of income and marital status—an assumption that can differ from reality.

Open the NHANES body-measures file ↗
Open the demographics and examination-weight file ↗
Open the published CDC validation table ↗

The optional body mass index (BMI) filter

The BMI-below-30 filter uses sex- and age-specific measured obesity prevalence from CDC/NCHS NHANES tables. Body mass index (BMI) is a clinical threshold used for one optional model input, not a measure of attractiveness, health, character, or worth.

Like height, it is applied independently because no source in the model jointly observes every selected trait. The source-input sensitivity range converts the table's published standard errors into approximate 95% normal ranges for obesity and then takes their non-obese complements.

Open the CDC prevalence tables ↗

The optional current-cigarette filter

The “no current cigarettes” filter uses 2024 National Health Interview Survey prevalence by age: 3.4% for ages 18–24, 10.5% for ages 25–44, 13.3% for ages 45–64, and 7.8% for ages 65 and older. The calculator applies the complementary non-current-smoking share to each exact-age Census cell.

This is a separately modeled national marginal rate. It is not jointly observed with income, education, marital status, height, or BMI; it is not state-specific; and it does not measure vaping, cannabis, former smoking, or compatibility. The source-input sensitivity range uses the complements of the published 95% confidence intervals for current cigarette use.

Open the CDC/NCHS 2024 NHIS table ↗

How much the estimate can vary

Two separate ranges show two different kinds of uncertainty. They do not predict dating outcomes or combine into one final confidence range.

When any modeled filter is active, the data receipt displays two deliberately separate ranges. The source-input range keeps the calculator's independence formula but varies the CDC inputs: published BMI standard errors, published cigarette-use confidence intervals, and an approximate Wilson interval around the fitted height-tail rate using effective sample size. Multiplying several input ranges does not create a final 95% confidence interval, so the interface does not label it one.

The dependence envelope asks a different question: how much could the result move if the Census-qualified group and the separately modeled health traits overlap differently than independence assumes? For each age cell, it applies the Fréchet–Hoeffding intersection bounds—at least max(0, sum of marginal shares − number of modeled conditions) and at most the smallest marginal share—then aggregates the age cells. The envelope can be wide because the source datasets do not identify the true joint overlap. It is a structural sensitivity bound, not a probability forecast or sampling interval.

Neither range includes ACS sampling error. The checked-in client tables retain the Census person-weight point estimates but not all 80 replicate-weight totals needed for design-based margins across arbitrary age ranges. Exact state ranks therefore remain withheld, and nearby point estimates should not be treated as statistically ordered.

State and city boundaries

Each state report uses its own weighted ACS PUMS records. Washington, D.C. is included. State selection changes the Census-derived age, income, education, and marital-status pool; the national CDC physical-trait distributions remain the model input.

Exact city counts are intentionally absent. PUMS identifies states and Public Use Microdata Areas, not every exact city. A future city estimate must use a separate geographic model and state its uncertainty.

Dating-spend survey

The spend section reports a 2026 Bank of America Better Money Habits survey conducted online by Ipsos among 1,133 U.S. Gen Z adults ages 18–29. The survey was weighted to national benchmarks and reported a sampling margin of error of plus or minus three percentage points.

The 51% zero-spend result covers the full surveyed population, including respondents who were not actively dating. It is not transaction data or a count of active daters.

What each source can tell us

These sources answer different questions. The site keeps their evidence types, units, and denominators separate instead of combining headline percentages into a synthetic dating statistic.

Why these percentages cannot simply be multiplied: each source can use a different population, question, unit, time window, and survey design. Multiplication would assume the traits overlap independently inside one shared denominator, which these sources do not observe.

When a claim crosses these boundaries, the page must name the new denominator and mark the result as modeled, cannot confirm, or needing a live source check.

How to read a result

The full group used for each percentage is the selected age-and-sex group in the selected place. “About 1 in N” restates that estimated share in everyday terms. When too few height records support an exact rate, the result shows an upper limit instead. When no matching adjusted Census records remain, the calculator does not show a one-in rate. This does not prove the real population is zero. Estimated counts can include part of a person because survey records represent many people. They are not a list of individuals.

ACS results shown here are person-weight point estimates. The data receipt shows health-input uncertainty and cross-dataset dependence sensitivity separately. Neither is a confidence interval for the final result. Pool Index rows use point estimates for direction only. Exact ranks are withheld because nearby differences may not be meaningful.

The Pool Index combined view is the geometric mean of the men’s and women’s qualifying group shares. It describes the two estimates together. It does not measure equality or dating success.

A page that asks about women leads with the women’s estimate. A page that asks about men leads with the men’s estimate. The other card uses the same filters for context. Neither card implies preference, compatibility, attraction, or intent.

A smaller estimated share is not a judgment. The model does not estimate mutual attraction, orientation, personality, intent, distance, app activity, or the probability of a real relationship.

See the separate expectations study protocol for preference data.

How we show what each setting changes

The setting comparison recalculates the estimate several times. Each check temporarily turns off one active setting while it keeps the place, age group, source group, and every other setting unchanged. The reported change shows how much that one setting affects the estimate.

These effects overlap. Many settings can describe the same people, so their percentages must not be added together. Age is not ranked because it defines the starting group.

If the current or comparison result lacks enough source records for a stable estimate, the exact impact stays hidden. The source label stays attached to each result: relationship status, education, and income are observed together in ACS records; height, BMI, and current cigarette use remain independently modeled from CDC/NCHS estimates.

How one setting changes the estimate

This comparison changes one eligible setting by the smallest available step—one inch of minimum height, $5,000 of minimum personal income, one education threshold, or turning off an optional BMI or current-cigarette setting. Every other setting stays fixed.

The reported change is relative to the current estimate for each source group. The same one-step change applies to men and women, so the two percentages can be compared directly. Relationship status is not included because changing it would change the legal-status group being described.

How state comparisons work

State comparison pages keep the example filters fixed and change only the state PUMS group. The page highlights the larger displayed share for each source group. It shows estimated headcount as context, not as a ranking rule.

The site publishes a selected set of useful regional comparisons rather than every possible state pair. Each page links to both full state reports and keeps the same source and model limits.