Who pays for these studies?

I'm not in the habit of fisking studies based on the abstract alone, but I'll make an exception in this instance.

Overestimation of Peer Smoking Prevalence Predicts Smoking Initiation among Primary School Students in Hong Kong

Purpose:

To investigate the relationship between perceived prevalence of smoking and smoking initiation among Hong Kong primary second- to fourth-grade-students.

Methods:

A cohort of 2,171 students was surveyed in 2006 and again in 2008. Students who perceived ever-smoking prevalence in peers as “none” or “some” were considered as correct (reference group), whereas those who perceived it as “half” (overestimation) or “most/all” (gross overestimation) were considered as incorrect.

Hmm. So if they perceived that none of their peers smoked, they were assumed to be correct. That may be true in the sheltered world of tobacco control, but for the rest of us that should be classified as an 'underestimate'. Except there isn't an 'underestimate' option available in this study, which leads me to think that it isn't very well designed.

Results:

At baseline, overestimation was found to be cross-sectionally associated with ever-smoking. At follow-up, 7.2% of never-smoking students with incorrect estimation at baseline had started smoking, which was 79% (95% confidence interval: 3%–213%), greater than that of 3.7% for those with correct estimation. Among the never-smoking students with incorrect estimation, subsequent correct estimation was associated with 70% (95% confidence interval: 47%–83%) lower risk of smoking initiation compared with persistent incorrect estimation.

Regardless of whether these kids' estimates are right, it's fair to assume that those who said 'most' had more friends who smoked than the ones who said 'some' or 'none'. And since having friends who smoke is a major predictor of smoking initiation, that—not the overestimating—is the reason they start smoking. The ones who said 'all' would, of course, be liars having a laugh at the researchers' expense. Given that the subjects are schoolchildren I believe, and hope, that there were many of them.

Conclusion:

Overestimation of the prevalence of peer smoking predicted smoking initiation among children. Interventions should be carried out to evaluate whether correcting children's overestimation of peer smoking could reduce smoking initiation.

Rubbish. The conclusion is that if your friends smoke, you're more likely to smoke yourself. But I think we knew that already, didn't we?