The Pool Report
How inputs become an estimate
  1. Census records

    The Census counts age, income, education, and marital status in the same records.

  2. Separate survey inputs

    You can add height, body mass index, and cigarette filters. We model them with other survey data.

  3. Population estimate

    See the estimated share that meets your filters.

On this page
  1. Who the calculator counts
  2. What Census records measure together
  3. Height is modeled from measurements
  4. The optional body mass index (BMI) filter
  5. The optional current-cigarette filter
  6. How much the estimate can vary
  7. State and city boundaries
  8. Dating-spend survey
  9. What each source can tell us
  10. How to read a result
  11. How we show what each setting changes
  12. How one setting changes the estimate
  13. How state comparisons work

Who the calculator counts

The calculator estimates how many adults ages 20–69 match your filters in the selected state or the U.S. The relationship filter means not currently married, not necessarily unpartnered.

It starts with the 2024 American Community Survey (ACS) Public Use Microdata Sample (PUMS). Census weights let each survey record represent people in the wider population.

ACS includes people in homes and group living settings. CDC health surveys cover civilians outside institutions, so the source populations differ slightly.

What Census records measure together

Age, annual personal income, education, sex, state, and marital status come from the same person’s record. These filters are counted together.

The source’s male and female sex categories appear as men and women in the calculator. ACS does not record gender identity or sexual orientation.

Income and education rules

We adjust personal income (PINCP) to constant dollars with ADJINC before applying a threshold. Education follows SCHL codes: a diploma or equivalent for high school, any postsecondary attendance for some college, and the named degree or higher for bachelor’s and graduate options.

The source generator groups these records into cells and is versioned with the site.

Height is modeled from measurements

Height comes from measured NHANES records collected from August 2021 to August 2023. We fit a height curve for each sex and ten-year age group, using CDC examination weights.

The curve estimates the share above each height cutoff. If fewer than five effective observations support the result, we show “1 in X or rarer” instead of an exact estimate.

Height is applied separately from Census traits such as income and marital status. Their real overlap may differ from this assumption.

Height model and validation

Model 2026.07-evidence-v3 uses normal distributions based on weighted means and standard deviations. Tests compare them with published CDC 5th–95th percentile points. Cutoffs range from 4′8″ to 6′6″.

For sparse results, an approximate 95% Wilson upper bound uses effective sample size. This is a stability check, not a full survey confidence interval. It does not cover every survey design feature or uncertainty in the curve.

Height measurements · Survey weights · Published CDC table

The optional body mass index (BMI) filter

The BMI-below-30 filter uses measured CDC/NHANES obesity rates by sex and age. Body mass index (BMI) relates weight to height. It is not an individual health assessment.

We apply this rate separately from the other traits. The source-input range uses approximate 95% ranges derived from the table’s standard errors.

CDC BMI tables

The optional current-cigarette filter

The 2024 National Health Interview Survey reports current cigarette use of 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.

We use the share who do not currently smoke for each age group. These national rates are applied separately from other filters. They do not cover vaping or cannabis. The source-input range uses the published 95% confidence intervals.

CDC cigarette-use table

How much the estimate can vary

Results with health filters show two separate ranges. Neither is a confidence interval for the final estimate.

The source-input range varies the CDC health rates while keeping the calculation’s assumption that traits overlap independently.

The overlap range allows the health traits and Census-qualified group to overlap differently. It can be wide because the sources do not measure every trait in the same people.

Neither range includes ACS sampling error. The client data do not retain all 80 replicate weights needed to calculate it. We therefore withhold exact state ranks.

Range calculations

The source-input range uses BMI standard errors, cigarette-use confidence intervals, and approximate Wilson height intervals. Multiplying input ranges does not produce a final 95% confidence interval.

The overlap range, also called the dependence envelope, uses Fréchet–Hoeffding bounds in each age cell: at least max(0, sum of marginal shares − number of modeled conditions) and at most the smallest marginal share. We then combine age cells. This describes possible overlap, not sampling error.

State and city boundaries

Each state and Washington, D.C. use their own Census records. Height, BMI, and cigarette rates remain national inputs. We do not provide city estimates because PUMS does not identify every exact city.

Dating-spend survey

Bank of America’s 2026 survey, conducted by Ipsos, included 1,133 U.S. adults ages 18–29. It was weighted to national benchmarks, with a reported margin of error of ±3 percentage points.

The 51% who reported no dating spending includes respondents who were not actively dating.

What each source can tell us

Census records combine age, income, education, and marital status. CDC supplies separate health estimates. Pew asks about self-reported relationship status. The Bank of America/Ipsos survey asks about dating spending.

These sources use different groups and questions. Their headline percentages cannot be combined as if they measured the same people.

How to read a result

“About 1 in N” restates the estimated share of your selected age, sex, and location group. When no matching Census records remain, we withhold a one-in rate; that does not prove the population is zero.

Weighted counts are estimates, not a list of people. Small differences between states may not be meaningful.

The combined Pool Index uses the geometric mean of the men’s and women’s shares. It summarizes both estimates, rather than their equality.

How we show what each setting changes

We turn off one active setting at a time and keep everything else fixed. The difference shows that setting’s effect on the estimate. These effects overlap, so do not add them.

Age defines the starting group and is not ranked. We hide exact effects when either result has too few supporting records.

How one setting changes the estimate

We lower one setting by its smallest step: one inch of height, $5,000 of income, one education level, or an optional BMI or cigarette filter. We apply the same change to both groups.

Relationship status stays fixed because changing it would change the group being counted.

How state comparisons work

State comparisons keep the example filters fixed and change only the state records. They show each group’s estimated share and headcount, with links to both state reports.