How can a small group reflect the diversity of the district?

01 The question it answers

At a community meeting, typically 40–100 people decide on matters that affect tens of thousands of residents. It is a fair question: why should we believe that this handful of people represents anyone but themselves? The answer lies in the way they are selected. If the participants are drawn so that the group mirrors the whole population along the important characteristics, then the small sample reflects the composition of the larger whole with surprising fidelity — just as the taste of an entire pot can be judged from a single spoonful of a well-stirred soup.

Stratified random sampling aims at exactly this: creating a "mini-community" that reproduces the district or town in miniature. It is not the loudest or the most invested who gather, but a group in which the many walks of life are present in roughly the same proportions as in reality.

02 Why isn't pure randomness enough?

At first it might seem obvious to simply draw 60 people at random from the population register and be done with it. Pure randomness is indeed free of bias — but it is unreliable on a small sample. In a group of sixty, chance can easily "slip": you may end up with disproportionately many older people or too few young ones, too many men or barely a handful of residents from the outskirts. On a large sample this would even out, but on a small one it can cause serious distortion.

Moreover, not everyone selected at random agrees to take part, and the refusals are not random either: certain groups systematically say yes less often. If we do not correct for this, the final group is skewed by willingness to participate. Stratification remedies exactly these two problems.

03 How does it work, step by step?

In practice the process usually breaks down into three stages.

• Invitation. The organisers send invitations to several thousand — typically five to ten thousand — residents chosen at random from the official population register. The large number is needed because only a fraction will respond.
• Registration. Those interested reply and provide a few basic demographic details about themselves: age, gender, level of education, place of residence and often other characteristics.
• Stratified draw. The final participants — typically 40–100 people — are drawn from among the applicants, but not blindly: the group's composition is made to mirror the whole population along the chosen criteria.

The last step is the key. The draw is still random, but framed by "quotas": if 20 percent of the population is over 65, then roughly 20 percent of the participants will be too. This way the purity of randomness and representativeness are secured at once.

04 Along which criteria do we stratify?

The criteria for stratification are determined by the topic and the local circumstances, but a few characteristics appear almost always. Age, gender, level of education and place of residence (for example a part of the district or town) form the basic set. Depending on the nature of the decision, others may join them: for a transport question it may matter whether someone travels by car, public transport or bicycle; for a housing matter, whether they own or rent.

The aim is not to cover every conceivable characteristic — that would be impossible and pointless. A few well-chosen characteristics relevant to the decision are enough for the group to reflect the community credibly.

05 Why the lottery, exactly?

Random selection — sortition — carries an important message in itself. In principle it expresses that anyone could have been chosen: participation is not a question of merit, influence or volume, but of pure chance. This reaches back to the ancient ideal of democratic equality, and that is precisely why it lends strong legitimacy.

It also has a practical advantage: the draw counters the bias of self-selection. It is not those already strongly affected by or interested in the matter who decide, but a balanced cross-section. Many become open-minded for exactly this reason: a participant chosen by lottery does not arrive as the representative of a camp, but as an ordinary resident willing to genuinely think the question through.

06 How is it different from an opinion-poll sample?

It is a fair question: if pollsters also take a representative sample, what is new here? The difference lies in the purpose and the scale. An opinion poll's sample serves to question people in large numbers, briefly, and to measure statistically accurate proportions — "what percentage of residents think this". The sample never meets, never talks, never deliberates; the answer is a snapshot of a momentary opinion.

The community meeting's sample, by contrast, serves not measurement but thinking together. It is smaller in number — since it has to work together over several days — but in return it is given time, information and dialogue. Stratification here serves not statistical precision but the aim that the group sitting in the room credibly reflects the diversity of the community, so that the deliberation is not one-sided either. The goal is therefore not to find out "how many" think something, but that a group mirroring the community can give a well-founded recommendation.

07 An example from practice

Domestic and international assemblies illustrate the scale of the method well. In a typical local process the organisers send invitations to several thousand — often five to ten thousand — residents, so that from among the applicants they can finally assemble a group of a few dozen, typically 40–100 people. International examples show a similar order of magnitude: the Irish assembly worked with 99 people, the French climate convention with 150, and city assemblies typically with 50–60. The common principle is the same everywhere: a large enough pool of invitees from which a manageable body mirroring the population is assembled by stratified draw.

08 What to watch out for?

The method only works if it is applied carefully. The number of invitees should be large enough that a balanced group can genuinely be formed from among the applicants. The barriers to applying — lack of time, distrust, digital limitations — must be deliberately lowered, for example through compensation, flexible scheduling and personal outreach; otherwise the very people usually left out will be missing from the applicants too. Finally, the criteria for stratification must be recorded in advance and transparently, so that the process remains verifiable and credible afterwards.

09 Summary

Stratified random sampling is the method that makes it possible for a few dozen people to credibly represent an entire community. Randomness ensures freedom from bias, and stratification ensures that the small group genuinely reflects the population. This technique is one of the cornerstones of deliberative participation — but on its own it is not enough: participants also need good information to make their decision. The next piece is about what makes background material balanced.

Sources

• Community Meeting – How does it work? · kozossegigyules.hu
• InnoK Knowledge Base – The model of deliberative democracy (James S. Fishkin) · urbanlab.hu
• OECD (2020): Innovative Citizen Participation and New Democratic Institutions · oecd.org
• newDemocracy Foundation: A framework for developing a public deliberation · newdemocracy.com.au
• Sortition Foundation: Stratified random selection · sortitionfoundation.org