Marijuana selection

You're going to mess up any estimate of the negative effects of using marijuana if you do not control for that the kids who try marijuana differ systematically from the ones who do not try marijuana.

If, as you might expect, trying drugs earlier correlates with things like novelty-seeking, higher risk tolerance, and perhaps less conscientiousness, and if you don't have particularly good ex ante measures of those things, you're going to confound the effects of early marijuana use with the effects of being the kind of person likely to try marijuana early.

Three years ago, Ole Rogeberg critiqued the Dunedin Longitudinal Data people for having inadequately controlled for these kinds of selection effects. Dunedin's response was not particularly good. I'd noted it here. There are big problems when researchers use secret data.

Ole's critique was published here; Dunedin (Moffitt, Poulton et al) replied here; Ole's rejoinder is here. 

It seems those selection problems are large indeed.

Now, in the first study of its kind, scientists have analyzed long-term marijuana use in teens, comparing IQ changes in twin siblings who either used or abstained from marijuana for 10 years. After taking environmental factors into account, the scientists found no measurable link between marijuana use and lower IQ.

“This is a very well-conducted study … and a welcome addition to the literature,” says Valerie Curran, a psychopharmacologist at the University College London. She and her colleagues reached “broadly the same conclusions” in a separate, nontwin study of more than 2000 British teenagers, published earlier this month in the Journal of Psychopharmacology, she says. But, warning that the study has important limitations, George Patton, a psychiatric epidemiologist at the University of Melbourne in Australia, adds that it in no way proves that marijuana—particularly heavy, or chronic use —is safe for teenagers.

The first quoted link goes to a paper that doesn't seem yet available. The second one controls for a broad array of ex ante characteristics and finds no effect of cannabis use. From its abstract:

A series of nested linear regressions was employed, adjusted hierarchically by pre-exposure ability and potential confounds (e.g. cigarette and alcohol use, childhood mental-health symptoms and behavioural problems), to test the relationships between cumulative cannabis use and IQ at the age of 15 and educational performance at the age of 16. After full adjustment, those who had used cannabis ⩾50 times did not differ from never-users on either IQ or educational performance. Adjusting for group differences in cigarette smoking dramatically attenuated the associations between cannabis use and both outcomes, and further analyses demonstrated robust associations between cigarette use and educational outcomes, even with cannabis users excluded. These findings suggest that adolescent cannabis use is not associated with IQ or educational performance once adjustment is made for potential confounds, in particular adolescent cigarette use. Modest cannabis use in teenagers may have less cognitive impact than epidemiological surveys of older cohorts have previously suggested.

And I wouldn't take smoking there as causal either. It's rather more likely a marker for other unobserved risk-preferences that correlate with the outcome variable and with drug use.

Remember this when somebody uses "Won't somebody think of the children" as first response against considering drug legalisation in New Zealand.

There's a decent write-up here.