[bmm6o]

And 100 HN points to you for spotting the fallacy! Well done, but you don't really believe the suggestion that after controlling for all else, closeness to Canada ends up being correlated with educational outcomes and crime rates, do you? It's obviously a joke.

[gwern]

I think that if you really control for all else like ethnicity, it would go away, but to condition on that much would miss the point.