As long as you split the distribution uniformly into categories with the same number of people, it doesn't matter that it's not uniform by date.
There is no doubt that you can uniquely identify the world population with 33 bits, so I guess the problem is to do it with questions people can answer themself.