I'd say a sample of 5000 random phone numbers is more representative than individual comments from a community obsessed (in a positive way) with privacy.