Blog

Correlation, causation, both?

A study in 2015 revealed (or confirmed) that former high school athletes have a much greater chance (about 45% greater) of joining the ranks of upper management in later life than their less athletic peers. A useful finding for employers seeking high achievers. Maybe not so useful for parents of high school students.

Have these students ascended to greater managerial heights because of their participation in athletics? Or do those who gravitate to organized sports tend to exhibit attributes that will later serve them well as executives? Ah, the classic nature-versus-nurture, chicken-egg conundrum: Is it correlation (perhaps there are other factors that encourage career success) or causation (the lessons you learn on how to succeed on the field are applicable to the board room)? The study does not conclude causation, so we’re left to accept only the correlation.

Here’s another one. A recent study from the University of Pittsburgh found that social media can be anything but sociable. Millennials who use social media for at least two hours a day are twice as likely than lighter users to feel socially isolated. Does use of social media contribute to social isolation, or do young people who feel marginalized to begin with tend to find refuge in social media? Probably both, but how much of each? Again, the study is inconclusive.

Unfortunately, in an information age stoked by sound bites and Tweets, the distinction between correlation and causation can often get lost. With evidence of correlation alone, we can only imagine how many over-weaning parents have tried to convert their bookish offspring into student-athletes to improve their career chances.

Uncovering correlations may be helpful in predicting the future. Understanding causation is the first step to controlling it. Conflating correlation with causation, however, can lead ultimately to ineffective action and the production of unforeseen and potentially counterproductive consequences.

Which is why any sound research needs to understand the relationships between dependent and independent variables and how together they can predict, or control, or both, or neither.