Do these percentages take into account causation vs correlation? Apparently (surprisingly), it's still very difficult to prove causation without a laboratory experiment.

Perhaps people who are likely to exercise have a certain genetic disposition to begin with.