We built Median Strategies because too much public polling asks you to just trust the topline number. We would rather show our work.
We draw from a stratified voter file sample, weighted to reflect the actual composition of the electorate we are studying by region, age, gender, and past voting history, not just raw population figures.
We use a mixed-mode approach, combining live-caller phone interviews with text-to-web response, to reduce the coverage gaps that come from relying on a single mode alone.
For primary and midterm electorates, turnout is the hardest variable to model. We weight rather than hard-filter on turnout history, so lower-propensity voters are represented at a realistic share rather than excluded entirely.
We weight to targets for age, gender, race and ethnicity, education, region, and past vote or turnout history, set against the best available public benchmarks for each state and race.
Every release states its margin of error clearly, both for the full sample and for any subgroup we highlight.
Every release includes full regional and demographic crosstabs, not just a topline number, and we make raw toplines available on request for credentialed researchers and journalists.