Second, for each cluster the same number of residential addresses is randomly sampled from the municipal registries (self-weighted sample of individuals).