Sampling · 3 min read
Census-matched online samples: what they are and how to get one
A census-matched sample is a sample whose mix of people matches the population on a few key characteristics, usually age, sex, race and ethnicity, and sometimes education. It's one of the most useful tools for online research, as long as you're clear about what it does and doesn't fix.
Updated November 02, 2026
What census matching does
You set quotas so that each group gets places in proportion to its share of US adults, then fill them from an online panel. The benchmark usually comes from the American Community Survey (ACS), the Census Bureau's large annual survey.
What it doesn't do
A census-matched sample is still an opt-in sample, not a probability sample. Matching on age, sex and race doesn't guarantee the sample matches the population on things you didn't match, such as political interest or internet use. Many researchers still weight the data afterwards, for example by raking to several benchmarks, and say so in their methods.
Joint cells beat separate quotas
Matching each characteristic separately can still produce odd combinations, such as plenty of older people and plenty of Hispanic people but very few older Hispanic people. Quotas on joint cells, such as age by sex by race and ethnicity, keep those combinations in proportion. The trade-off is that small cells are harder and slower to fill.
Where the numbers come from
Published census tables rarely give every combination you need, so joint proportions are usually computed from the ACS Public Use Microdata Sample, weighting each person record by its person weight. Turning proportions into whole-number quotas uses a rule such as largest-remainder allocation, so quotas add up exactly to your sample size.
Practical tips
- Keep the number of cells reasonable for your sample size; very small cells add delay without much gain.
- Allow extra time for hard-to-reach groups to fill.
- Decide in advance whether you'll also weight, and on which variables.
- Report the benchmark, its year, the variables and how quotas were allocated.
How Populiva does it
Census-matched studies on Populiva split places across age, sex and race or ethnicity, optionally with education, in proportion to US adults. The targets come from the 2024 ACS 1-year microdata, and the signed provenance report for the study shows how many places in each quota were filled.
Run a census-matched study and get the quota report with your data.