Did you know that margarine consumption causes divorces in the state of Maine? There’s nearly perfect correlation of these two between 2000 and 2009 based on US Census data. 100% of the people who die were exposed to water in their life. The Chicago Bears football team had a winning season every year my wife and I lived there. Ridiculous, right?
You’ve heard it before: “Correlation does not imply causation.” Causation requires a this-then-that-follows sequence over time. Correlations may be quite strong, but not indicate causation – and bad decisions are costly. Yet leaders fall into the correlation trap all the time. We make decisions based on data available, and a strong desire to make sense of the world.
Three fundamental drivers for this trap:
We don’t have enough useful data. It’s easy to draw wrong conclusions from a limited amount of data, or think that something will be the same this time because it was last time. Useful data has both breadth and depth. Systems of any complexity have noise. Be particularly wary of assigning causation when you have only a small amount of data.
We want things to be simple. Let’s face it, life is messy. We’d like to have simplicity instead of complexity. Simple answers don’t exist for complex problems (but your leadership needs something better than “It depends”). You want simplicity on the far side of complexity. Don’t assume causation because it would be simpler. Form a hypothesis and test it.
We aren’t interpreting existing data correctly. It’s quite human to prefer one narrative over others, even in the face of conflicting data. Confirmation bias — when we pay attention only to elements of data which support our preferred perspective – can lead you into this-causes-that errors. You don’t have to be a trained statistician to interpret data correctly.
Leaders, check yourself: It is truly causation, or just correlation?