Stratification is a method for shaping a group so that it reflects the diversity of a wider population across chosen characteristics, such as age, gender, geography, or socio-economic background.
# Strata
These targets are the “strata.” Age bands, gender, region, sometimes education, sometimes income brackets, sometimes ethnicity, sometimes urban and rural. The details vary, and that variation matters, because each choice expresses a theory of what “representative” should mean.
Rather than selecting participants purely at random or allowing open self-selection, stratification sets explicit proportions for each category and then fills them, usually through random selection within each group. In democratic design, stratification is not about engineering outcomes, but about ensuring that who gets to speak and deliberate is not accidentally skewed by chance, convenience, or existing inequalities.
# The Fix
But Stratified Sortition is also a confession. It admits that modern deliberation usually needs consent. In most cases, participation is voluntary. People have jobs, kids, chronic illnesses, migration status complexities, fear of public exposure, and a thousand other reasons to say no.
That means the real selection pool is not “the whole population,” it’s “the people who can and will show up.” Stratification is partly a fairness tool, and partly a repair tool: it tries to correct the predictable patterns of who opts in.
There’s a subtle tension here. Purists sometimes want “pure sortition” because it feels cleaner. Practitioners want a group that is visibly legitimate to the public and resilient to criticism. Stratified Sortition is the compromise that has become the default, because legitimacy is not only about statistical purity. It’s also about whether the people in the room can plausibly claim to speak from within many lived realities, not just one.