Because there are much more families that do not have alcohol problems than there are families with an alcohol problem. Thus, if we assume for simplicity that the chances of developing an alcohol problem is purely random(*), the persons that deveolop alcohol problems will be distributed according to the ratio non-problem:problem. And since non-problem > problem, there will be more from a non-problem background than from a problem background. Quite simply.
(*) That’s obviously not quite true and an oversimplification, but for illustration purposes it is good enough.