Compare communities.
Understand circumstances.

Rates are estimates, not exact counts. Intervals show 95% confidence based on all 80 Census replicate weights. The difference uses paired replicates, including covariance when the groups overlap. A Bay Area comparison with one of its counties therefore does not treat the groups as independent.

Observed rates describe each group’s actual composition in the survey. Age-standardized rates ask how the comparison looks if both groups have the Bay Area’s outcome-eligible age distribution. They adjust age only, not other differences in people’s circumstances. Standardized rates do not have corresponding population counts.

These user-selected comparisons are exploratory; their confidence intervals are not adjusted for trying many comparisons. The report separately documents a fixed family of eight comparisons and a Holm adjustment.

English proficiency is self- or household-reported. It does not establish preferred service language, language-learning ability or the full range of languages someone knows. Language use, race, ethnicity and citizenship are different concepts.

SLP requires at least 100 eligible records and at least 10 records with and without the outcome. Estimates with relative standard error above 30% are marked imprecise and excluded from differences. Age standardization also requires at least 30 records and five outcome and non-outcome records in each age band. These are SLP analytical rules, not official Census publication thresholds.

An unavailable estimate does not mean no people have that experience. Some combinations cannot have the outcome by definition (for example, English-only speakers cannot meet the limited-English definition). Small communities remain part of regional totals. “Other” combines remaining languages and is not a single language community.