I’ve spotted the problem. Your sample size is too small. The world is quite a bit larger than “social media”.
It's kind of rare that a result is poor due to sample size - it's almost always the methodology that dooms it.
A typical example is the birthday "paradox" - where a sample size of just 23 **RANDOMLY** selected people have a greater than 50% chance of two people having the same birthday (and it's 97% once you reach 50) - and you scratch your head and dismiss it and tell me it can't be, it defies all logic. And a kid might answer "well there's more than that many in my class, and that's not true" and another might say "can't be - I have more than two dozen friends, and it's not so".
Except - those aren't RANDOM. A classroom of kids all born the same year isn't random. A selection of your relatives, friends and neighbors - isn't random. The participants of this forum - isn't random. Guy in the street stuff isn't random, and may well be intentionally or unintentionally skewed by the interviewer.
