Andrew Morral and his colleagues at RAND seem to have put a major hole in the “gateway theory” so beloved of the drug warriors. Their paper, which appears in the current issue of Addiction, isn’t yet up on the journal’s website, and when it is it will be available by subscription or one-time payment only. (Look for Vol. 97, Issue 12.) In the meantime, here’s a rather mealy-mouthed press release from RAND (eager to distance itself from any hint of supporting the l-g-l-z-t–n of dr-gs) and a pretty good UPI story. [Note that the DEA spokesperson utterly misses the point.]
I predict the study won’t make much of a splash, despite the mention of it in Bill Keller’s New York Times column on Saturday, for two reasons: (1) it’s methodologically complicated (using a simulation model) and (2) it reaches the politically incorrect conclusion.
Here’s my while-I-stand-on-one-foot version of the story:
1. There is a strong observed correlation between marijuana use and use of other drugs. The “risk ratios” (the conditional probability of, e.g., using cocaine given marijuana use divided by the conditional probability of using cocaine given no marijuana use) are in the double digits.
2. That has led people to postulate a “gateway effect” by which marijuana use causes the use of harder drugs (in the sense that, other things equal, a change in external conditions leading some people to use marijuana who otherwise wouldn’t have will increase the proportion of that population that goes on to use, e.g., cocaine).
3. The alternative assumption is that the correlation reflects an underlying characteristic of people that leads them both to take marijuana and to take cocaine.
4. The difference has important implications for policy choice. If the correlation reflects merely differences in underlying propensities for illegal mood-alteration, then changes in marijuana availability, price, or consequences that lead to changes in the prevalence of marijuana use won’t have any impact on the level of cocaine use. If the correlation reflects instead a real “gateway effect,” then loosening up on marijuana policy now would tend to increase cocaine use later. (That the big increase in youthful marijuana use in the 1990s hasn’t yet produced any echo in cocaine use is encouraging, but not dispositive, on this point.)
5. Morral et al. tried to determine whether that “propensity” interpretation can explain the observed correlation, or whether a “gateway effect” is needed.
The succeeded in building a model that assumed there was no gateway effect, but which still fit the observations.
7. Ergo, the gateway effect is a superfluous hypothesis, not needed to explain the phenomena, and (by Occam’s Razor) ought to be discarded.
8. Given the design of the study, it could not “disprove” the existence of a gateway effect. But the finding removes any need to hypothesize such an effect in order to explain what actually goes on in the world. Until someone comes up with new observed phenomena that can’t be explained in this way, or shows that the underlying propensity assumed by the study either doesn’t exist or doesn’t have the assumed distribution, the gateway hypothesis must be classed with what Mark Twain called “vagrant opinions, living without visible means of support.” (Unfortunately, the social safety net underlying vagrant opinions of this type is quite strong, with no work requirement or five-year time limit.)
Morral’s results don’t imply any particular set of marijuana control strategies. (The Reuters story illustrates confusion on this point by calling the gateway theory “a basic principle of U.S. anti-drug policies,” as if even a well-supported causal hypothesis could by itself be a principle of action.) Marijuana drug causes enough damage on its own account so that one could reasonably oppose commercial legalization (as I in fact do) or even support full prohibition (as I used to do and might do in the future) without making any “gateway” claim at all.
But I doubt that many people or organizations on the pro-drug-war side of the debate will want to make that rhetorical shift. That side had made too heavy an investment in the gateway idea, and not been nearly careful enough in marking it as a plausible causal mechanism rather than an established fact. I predict that those who find the results inconvenient will ignore them, lie about them, or slander the authors, rather than sending a favorite argument to the recycling bin.
And I should have referenced Rob MacCoun’s clear-headed exposition of the seven different things “gateway” might mean.