Marijuana as a gateway drug

One of the claims made by the Office of National Drug Control Policy and others who oppose relaxing our drug laws is that marijuana is a gateway drug; that is, using marijuana may lead to using harder drugs. See for example page two of this report (.pdf file), or here about the theory applied to alcohol and tobacco.

If you do a study and find that most of the people using harder drugs like cocaine and heroin started off using marijuana or alcohol, then can you conclude that the alcohol and drug use is the reason for the harder drug use? Is this enough to establish a cause and effect relationship?

Probably not, as this explanation of research from the Rand Corporation explains (which is also mentioned in a story at Reason magazine).

This example illustrates an important statistical concept of "correlation without causation". Just because one thing follows another doesn't mean that the first caused the second. It's pretty clearly true that many heroin users also use(d) cocaine, but it's quite different to say that there is a causal link. What's happening is that there is a missing factor (such as a predisposition to using drugs) that explains both behaviors. I'm sure you'll be able to find other examples of this statistical concept. One that comes to mind is whether going to a top university leads to higher salaries. Is it the going to the Prestigious U. that is the reason or are exceptionally bright people going to these schools and exceptionally bright people get higher salaries anyway, which means that "exceptional brightness" is the missing factor?