xkcd offers a great example of the timeless conundrum: correlation or causation? The hatless stick figure uses data to suggest that using cell phones cause cancer, the hatted stick figure suggests that cancer causes cell phone use:
While it may seem ridiculous to suggest that cancer causes cell phone use, a lot of data is a bit more difficult to parse out. Does depression cause homelessness or does homelessness cause depression? Does a college education instill ambitiousness or does ambitiousness lead people to seek a college education? Does poor health result in poverty or does poverty result in poor health? Without a controlled experimental or longitudinal data that surveys the same people repeatedly over time, causation simply can’t be established. But it sure is tempting to guess.
Btw, if you’re teaching Methods next year, as I am, you might find useful our 80+ posts on methods and use of data. Good times!
Thanks to Paul Kingsley for reminding me about experimental methods.
Lisa Wade is a professor of sociology at Occidental College and the co-author of Gender: Ideas, Interactions, Institutions. You can follow her on Twitter and Facebook.
