Confusing correlation with causation is natural. We all tend to fall into the trap.
I. Alcohol Sales Outlets
II. Alcohol Merchandise
III. Ads for Alcohol
IV. Eating Together
The classic example is the correlation between high ice cream sales and drownings. Or the sale of sunglasses. But ice cream sales don’t cause either drownings or the sale of sunglasses.
The cause of these things is clearly warm weather. It’s not caused by high ice cream sales!
I. Alcohol Sales Outlets
Similarly, people often observe that alcohol consumption tends to be higher in some areas. Specifically, in those places with many liquor stores and other alcohol sales outlets. They follow this by a call to reduce the density of such outlets. They say this would reduce consumption.
However, the assumption that alcohol outlets cause high consumption appears to be wrong. Don’t stores locate where sales are higher?
The economist Dr. Steven Levitt relates this folktale. The czar discovered that the most disease-ridden province in the country had the highest concentration of doctors. To reduce the high disease rate the czar had all the doctors in that province killed.
Be sure to visit Spanking and Later Alcohol Abuse. It’s another great example of confusing correlation with causation!
Some would have us believe that the alcohol sales outlets are causing the drinking rate. They would have us follow the lead of the misguided folktale czar. Fortunately, there would be no executions. Yet they would have us reduce the number of sales outlets. That’s just as illogical.
Reducing the number of alcohol sales outlets is a “feel good” approach doomed to failure.
II. Alcohol Merchandise
“Alcohol Merchandise Encourages Underage Drinking.” That headline, was uncritically carried by major news sources. The widely-printed article said the sale of alcohol-related merchandise to those under 21 was “a wake-up call” for the nation.
What did this study of a small sample of young people from Vermont and New Hampshire reveal? Simply that young people who drank were more likely to have shirts and other merchandise with alcohol brand logos.
Of course, religious people are more likely to wear crosses, Stars of David or other religious symbols than are non-believers. Wearing religious items doesn’t cause people to become religious. And wearing a shirt with a beer label doesn’t cause people to begin drinking alcoholic beverages.
The research also found that “sensation-seeking” is associated with wearing alcohol merchandise. It’s also correlated with drinking, having peers who drink, having tried smoking, and doing less well in school. The enduring trait of sensation-seeking is more likely to cause drinking than wearing a cap showing a beer brand.
A headline of “Sensation-Seeking Personality Encourages Underage Drinking” wouldn’t grab attention. But it would be much closer to the truth.
It’s another example of confusing correlation with causation.
III. Alcohol Ads
In making this leap of faith, they ignore evidence to the contrary. For example:
- People who want to buy a new car tend to pay more attention to auto ads. People who want to buy a new house tend to pay more attention to real estate ads. And those who want to drink probably pay more attention to alcohol ads.
- Expenditures for alcohol ads have little or no relationship to alcohol consumption.
Research on this subject has been conducted for decades. And by governments, health agencies, and universities around the world. The result? There’s no good evidence that alcohol ads cause non-drinkers to begin drinking. Nor drinkers to consume more.
Then why do alcohol beverage producers advertise? They advertise to increase their market share. Both research and experience has shown that effective advertisers can increase a producer’s share of the market. It gains share at the expense of others, who lose some of their share.
However, research consistently shows that ads do not increase overall alcohol consumption.
The authors of the attention-grabbing report need to go back and take Research 101.
IV. Eating Together
A nation-wide public-service campaign urged parents to eat dinner with their children. This was to reduce youthful drug use. The advice was based on a report. The report supposedly showed that eating dinner as a family reduced risk of drug use. Specifically, it reduced it by half among young people.
“There is no more important thing a parent can do” to reduce the chance that their children will use drugs. So insisted anti-drug leader Joe Califano. He said that it “ is the key to ridding our nation of the scourge of substance abuse.”
Eating together as a family probably has many benefits. But what of the claim that doing so reduces drug use risk by 50%?
Study Did Not Show
People should see red flags everywhere. Indeed, the report could well serve as a case study in how not to conduct research.
For starters, the samples of parents and young people is not at all representative of U.S. families. The researchers began with over 37,000 phone numbers. They dropped one-third because of language barriers, no one answered the phone, and other reasons. Of the remaining phone numbers, 9,000 were people who refused to answer questions, and about 1,000 “were cut off.“
The company that conducted the survey admitted that it had “a very low response rate.” Yet nothing was done to address this serious problem.
The report concluded that frequently eating together as a family reduced drug risk by 50%. But it failed to consider the effects of age. Seventeen-year-olds are much more likely to use drugs than are 12-year-olds. They are also much more likely to eat apart from other family members. Ignoring this important and obvious fact makes the findings meaningless.
What the Study Did Find
However, science writer Carl Bialik asked the survey company to conduct a simple analysis. That is, to examine the effects of age. The result? Not surprisingly, “age correlated more strongly with risk than did family dinners.”
The investigators also failed to take into account family socioeconomic status or geographic region. And the list goes on and on.
Another serious problem is that the researchers didn’t actually compare family dining with drug use. Instead they compared dining behavior with a “drug risk score” that they made up. The risk score was based on such things a whether or not respondents said their friends use drugs.
The investigators actually asked respondents if they used drugs themselves. However, that information was not used for reasons unknown. Perhaps it didn’t support the reports conclusion. But who knows?
No Peer Review
The report was self-published by the sponsoring group. Thus, it avoided the scrutiny of peer review. And that’s required for publication in scientific and journals.
Reports that bypass the peer review process may use faulty sampling techniques, improper statistical analyses, draw unwarranted conclusions, and make unsubstantiated assertions. In the absence of peer review the public is completely unprotected. Special interests groups can, and often do, pass off shoddy and even intentionally deceptive reports as legitimate scientific research.
Eating dinner together regularly is probably a reflection of strong family ties. Of a desire to communicate. And of a generally well functioning family. In short, it’s almost certainly a result rather than a cause of such functioning. Therefore, a dysfunctional family that decides to eat together regularly will probably fail to prevent abuse.
So what’s the harm in urging people to eat together? Nothing. Unless it misleads people into thinking that by simply by doing so they can solve the problem of substance abuse.
This is a great example of confusing correlation with causation.
V. Resources: Confusing Correlation with Causation
- Alcohol Use and Abuse: How to “Lie” with Statistics
- Underage Drinking
- Alcohol Advertising
- How Accurate are Statistics on Alcohol Problems
- Alcohol, Law & Policy
- Brignell, J. Sorry, Wrong Number! The Abuse of Measurement. London: Brignell, 2000.
- Huff, D. How to Lie with Statistics. New York: Norton, 1993.
- Milloy, S. Science without Sense: The Risky Business of Public Health Research. Washington: Cato, 1995.
- _______. Junk Science Judo. Washington: Cato, 2001.
Note on Confusing Correlation with Causation
This website provides information. It usually makes no suggestions. However, in this case, it urges readers to avoid confusing correlation with causation. Indeed, that is the purpose of this web page.