Correlation vs. Causation and the phrase “Attributable To”| Vol. 2 / No. 45.2

Is not getting a degree as bad for you as smoking? Well, it depends on what you mean.

***

Last week a paper published in the journal PLOS ONE came to my attention. It had actually been published back in July, but it popped up in my feed because of an article in Vox. The article, in a predictably over-the-top fashion, is titled “A Lack Of Education Could Be As Deadly As Smoking. Here’s Why.” At this point I think I’m just going to start ignoring the titles of articles. They’re rarely written by the authors of the articles, and they’re often a sensational take on an already news-friendly (read: less accurate, but more readable for the general public) version of an often technical story.

The article itself centers around a paper, titled “Mortality Attributable to Low Levels of Education in the United States,” and seems to be making a simple (common!) mistake in science reporting: mistaking correlation for causation. From the article:

“In 2010 alone, more than 145,000 deaths in the US could have been prevented if high school dropouts had finished their secondary education, three public health researchers argued in a recent paper. Getting college dropouts to graduate could have saved 110,000 more.”

On the surface that’s an absurd claim to make, and thankfully the author of the article walks it back a tiny bit it a little later: “It’s conventional wisdom that people with more education are also healthier. But it’s surprisingly difficult to untangle why that’s the case, because it’s not clear whether it’s the education or the degree itself that makes the difference.” (It’s more than conventional wisdom, actually. And have you ever noticed that articles that include the words “here’s why” in the title are rarely able to deliver on that promise?)

And you’d think that the Vox article is misunderstanding the PLOS ONE article, but you’d be wrong. That’s really pretty close to what they’re claiming: “The estimated 145,243 deaths in 2010 attributable to individuals having less than a high school degree rather than a high school degree or GED, based on educational disparities in mortality in the 1945 birth cohort, is comparable to the estimated number of deaths that could be averted if all current smokers had the mortality rates of former smokers.” (9-10). [emphasis mine]

But the study doesn’t do much to back up the claim of causality that “attributable to” at first seems to imply. In fact the paper frequently uses the term “attributable to” — with, I can only assume, a great deal of emphasis on the –able. You are indeed able to attribute these deaths to poor education. But I wouldn’t recommend it, because chances are good that deaths correlated with educational attainment are just one part of the large systemic issue called poverty.

What the study did was find a correlation. People with less education died more and sooner than people with more education. But they didn’t control for socioeconomic status (which we know affects educational attainment): “Potential confounders include childhood health, cognitive ability, genetic predispositions, and childhood socioeconomic conditions such as living with low-educated, unmarried, or impoverished parents in early life. At present, no available data allow us to incorporate all potential confounders.” The authors of the study then go on to claim that since “the association between educational attainment and U.S. adult mortality has been robust to inclusion of such confounders when available” they therefore feel safe in asserting that “their inclusion would not fundamentally alter our findings” (10). Well, okay… but could you go into it a little more? No, apparently.

So, fine. If it does show some kind of causative effect, how might low education affect all-cause mortality? They don’t actually get to that, either, but we can talk about it here. It’s possible: low education might lead to more deaths because it perpetuates lower income, which correlates with poorer health outcomes (for a variety of reasons, including job insecurity and time restrictions, and the cost of medical insurance), higher stress (itself correlated with all-cause mortality), and so on. But low socioeconomic status affects educational outcomes, the stress of poverty affects cognitive ability, and if you’re poor you’re likely to stay poor.

I don’t doubt the correlation is real and not a statistical anomaly, but correlation isn’t causation, and it seems to me that we have plenty of data to suggest that educational attainment is just one part of a systemic issue, a series of interrelated problems that feed back into one another to eventually lead to higher all-cause mortality.

So, yes, attributable — as a part of a larger network of socioeconomic factors that effect health outcomes and, via that, mortality rates. But until we have more data, I’d stay away from even implying solely attributable, as the line from the Vox article seems to do. Or for that matter calling lack of education as bad for you as smoking. There’s just more to it.

***

Richard Ford Burley is a writer, library worker, and doctoral candidate in English at Boston College, where he’s studying remix culture and the processes that generate texts. In his spare time he writes about science, skepticism, and feminism (and correlation and causation) here at This Week In Tomorrow.