You keep saying that I have a sampling bias, but there really isn't any evidence for that. I'm sampling 100% of the population. You can't have a sampling bias when sampling 100% of the population.

It could be a spurious correlation, sure - but that'll go away as the amount of data increases.