Surveys like that are often designed to get a representative group. It's not like they call thousand people on random and keep their responses outright. After

Surveys like that are often designed to get a representative group. It's not like they call thousand people on random and keep their responses outright. After a while a certain demographic will be "full" and as such the effort will be to get responses for the groups not adequately represented yet. Or another way is to weigh the answers. If you get 90% responses from elderly and 10% from young, you cannot just use the average. But can use models to apply those groups' answers to how big the group is in the demographic of the country.