Study: Smart People Drink, Smoke, Use Drugs
By Ronald Bailey
Researchers find in a new study published in the Review of General Psychology that people with higher IQs are more likely to try out and enjoy mind-altering sutbstances than their dimmer compatriots. As the abstract reports:
Why do some individuals choose to drink alcohol, smoke cigarettes, and use illegal drugs while others do not? The origin of individual preferences and values is one of the remaining theoretical questions in social and behavioral sciences. The Savanna-IQ Interaction Hypothesis suggests that more intelligent individuals may be more likely to acquire and espouse evolutionarily novel values than less intelligent individuals. Consumption of alcohol, tobacco, and drugs is evolutionarily novel, so the Savanna-IQ Interaction Hypothesis would predict that more intelligent individuals are more likely to consume these substances.
Analyses of two large, nationally representative, and prospectively longitudinal data from the United Kingdom and the United States partly support the prediction. More intelligent children, both in the United Kingdom and the United States, are more likely to grow up to consume more alcohol. More intelligent American children are more likely to grow up to consume more tobacco, while more intelligent British children are more likely to grow up to consume more illegal drugs.
Based on data from the National Longitudinal Study of Adolescent Health. Psychology Today also reported:
The following graph shows the association between childhood intelligence, measured in junior high and high school, and adult alcohol consumption seven years later in the National Longitudinal Study of Adolescent Health (Add Health) data in the United States. The association is clear and nearly monotonic. The more intelligent Americans are in their childhood, the more alcohol they consume as young adults.
It is important to note that both income and education, as well as childhood social class and parents' education, are controlled in multiple regression analyses of these data from the US and the UK. It means that it is not because more intelligent people occupy higher-paying, more important jobs that require them to socialize and drink with their business associates that they drink more alcohol. It appears to be their intelligence itself, rather than correlates of intelligence, that inclines them to drink more.
I suspect that many H&R readers already thought that this might be the case.
Sign up for the OV Daily Newsletter
Sign up for the OV Daily Newsletter
Repeated studies in the US have shown the education and income are directly correlated with the consumption of alcohol and tobacco, meaning the less you make and/or the lower your education level, the more likely you are to consume alcohol or tobacco.
http://www.gallup.com/poll/127532/Income-Education-Levels-Combine-Predict-Health-Problems.aspx
http://www.niaaa.nih.gov/RESOURCES/DATABASERESOURCES/QUICKFACTS/ALCOHOLCONSUMPTION/Pages/dkpat17.aspx
This hasn't been correlated by IQ, but it certainly brings into doubt this study. Unless IQ is inversely correlated with education and income levels, which makes no sense.
I think this study is bogus.
I have not read the original study and so won't comment on whether or not it is "bogus", but there is no contradiction between the data you cited and this new research. What you describe would only be a problem if data could only have ONE correlate. It is entirely coherent, though, to say that X contributes to Z AND Y contributes to Z. eg
"Some people are overweight because they eat too many pizzas"
"Some people are overweight because they never exercise"
"Some people are overweight because both of the above"
No contradiction there, right? So why, then, see a contradiction in the identically formed (though obviously oversimplified)
"Some people smoke because they're poor"
"Some people smoke because they're smart"
"Some people smoke because both of the above"
Not only is this not a contradiction, but this is how influencers for ALL complex behaviors work!
---
The article specifies that the researchers controlled for education and income, among other things. So you can imagine a simplified data set like this:
Poor, smart person: 100 units of drugs and alcohol per year
Poor, less-smart person: 50 units of drugs and alcohol per year
Wealthy, smart person: 20 units of drugs and alcohol per year
Wealthy, less-smart person: 10 units of drugs and alcohol per year
---
Now this data is artificial and perfectly correlated, but you should see how real-world data could fit this pattern. Notice that intelligent subjects consume twice as many D/A units as their less intelligent counterparts when we ignore income differences. At the same time, notice that poor people consume 5x as many D/A units as their wealthier counterparts, when we control for intelligence.
The reason that one potential correlate is held constant while another is investigated is to ensure that researchers aren't overlooking a connection between intelligence and wealth that is so strong that it misrepresents the other (D/A) relationship. Insofar as the article reports it, they did their job and there's nothing wrong with the data on that superficial level. We don't really have the information to comment further without going into the published paper itself.
Hope this helps clear things up!
I did some research of my own, and this study smacks of a Type II statistical error. There are missing covariates/factors that might account for increased consumption of drugs and alcohol such as increased stress levels that students and adults with higher IQ experience due to achievement pressures. The initial results require much more study before this could be called conclusive. Also, the effect sizes are minuscule compared to the effect sizes of income and education levels.
Consequently, I understand the study and I believe the correlation is more likely to be a type II error.
We need to avoid drawing any conclusions from something like this. It seems like something building toward an excuse for bad behavior.
This study reminds me the studies that showed that kids who miss breakfast were more likely to do poorly in school. What was actually found was that poverty, parental involvement, and mother's education level were actually much more predictive of student performance. The way researchers discovered that the impact of breakfast was much less than thought was that giving the kids breakfast didn't improve test scores by a significant amount.
Right now, I would say any public funding for this type of research would be money poorly spent given the effect size.
One more followup. You provided two links in your first comment, in support of your suggestion that there is an inverse correlation between education level/income andndrug/alcohol use.
While the tobacco use data you liked to bears this out, the NIH data on alcohol consumption ( http://www.niaaa.nih.gov/RESOURCES/DATABASERESOURCES/QUICKFACTS/ALCOHOLCONSUMPTION/Pages/dkpat17.aspx ) directly contradicts the point you're trying to make with it. From the data provided (1992; 1987 correlations appear to be similar):
Abstinence - Higher education levels = less abstaining
---
67% of those with "some high school" education abstain from alcohol.
50% of those with a completed "high school graduate" level of education abstain
38% of those with "some college" education abstain
33% of college graduates abstain
Light drinking - Higher education levels = more light drinkers
---
16% of those with "some high school" education drink lightly
25% of those with a completed "high school graduate" level of education drink lightly
31% of those with "some college" education drink lightly
34% of college graduates drink lightly
Moderate drinking - Higher education levels = more moderate drinkers
---
12% of those with "some high school" education drink moderately
18% of those with a completed "high school graduate" level of education drink moderately
23% of those with "some college" education drink moderately
27% of college graduates drink moderately
Heavy drinking - There appears to be no or almost no correlation between education level and heavy drinking
---
6% of those with "some high school" education drink heavily
7% of those with a completed "high school graduate" level of education drink heavily
8% of those with "some college" education drink heavily
7% of college graduates drink heavily
Unless there's another way to read this data that I've missed, I think you misread this data as its opposite.
The Gallup data shows a strong correlation between lack of education and lack of income and smoking. However, the Gallup poll shows the opposite for alcohol.
Mea Culpe... I conflated the results of alcohol and tobacco.
How do I delete my other comments? Dang it!
I'm not about to get into an extended debate about a study whose data I can't see--if you have access and ability to Share it, I'd be happy to look, but I'm not about to buy the paper just to satisfy my curiosity--but a few quick things about the last thing you wrote:
- it is by no means widely accepted that stress levels and intelligence are correlated. Find three studies and they'll tell you three different things.
- it's been a long time since I've had to deal with this terminology, but I'm not sure how this falls into the category of "type ii error". Those are the false negatives, right? So unless I'm missing your frame of reference, wouldn't the type ii error here be if the researchers saw no correlation between intelligence and drug/alcohol use when really there was one? Please clarify if i'm misremembering/misunderstanding.
I'll repeat that, without seeing the actual research, I can't come to a confident opinion on this study. When you write "it seems like something building toward an excuse for bad behavior", though, it seems like your thoughts on this relationship were all sewn up before you ever looked at the research. That's the minefield of cognitive bias; when you take the position that research is dangerous if it leads to justifications/rationalizations for what you consider to be "bad" behavior, you're WAY more likely to come to the conclusion, as you have here, that there "just couldn't be" a relationship. Whether this particular study is good or bad, I worry from your tone and word choices that you dislike the IDEA of this relationship existing so much that you'd require an unreasonably high level of evidence to consider it as true,
I hope I'm wrong; that's not the stuff good science is made of!
I am stating that this is a type II because the effect size of other factors has been obscured by the covariates. It could also be called a type I error, meaning there is a correlation that is not related, meaning stress and not IQ is the cause of the correlation and controlling for stress without IQ could be a much stronger relationship.
The real problem with the study in my mind is the application of the Savanna-IQ hypothesis. This relationship and the focus of the study seems like a wild stretch. For example, the huge effect size of lack of education actually infers that *not* using alcohol or Tobacco is the true application of this hypothesis.
I find the whole chain of reasoning spurious and unsupported by this weak correlation and effect sizes.
I provide statistical models to clients around education and criminal recidivism risks, and I would be *fired* if I tried to a present a study like this.
Have you read the original paper, or are you speculating about whether the effect size has been ignored? I ask because the journalism on this research does assert that the researcher's accounted for the effect of education . I'd be curious to know why you disagree.
Also, please see my reply (coming in just one second) to your earlier post--I think that the data you presented earlier is problematic for your suggestion regarding the relationship between education and alcoholism, anyway.
It's fun to drink. Smart people understand this, dumb people don't. The study makes perfect sense to me.