Statistics for Skeptics Part 2 – Correlation vs. Causation

If you live in one of the counties around Edmonton, you’ve probably heard about the Alberta government’s plans to build some new high voltage power lines. In November, there was a protest staged by R.E.T.A and Strathcona County passed a resolution to oppose the construction. There are plenty of reasons for the uproar. Unfortunately, one of them is that people are worried that the power lines will cause cancer.

While that’s a topic that deserves an entire post of its own, it does present a good example of a part of statistics that the general public seems to miss: correlation.

A correlation is when two observed variables can be shown to have a statistical relationship. My favourite example of this was presented by my Sociology 100 professor, who had a plot showing ice cream sales vs. crime rates. There was a clear trend. As ice cream sales increased, crime rates increased. These two variables are strongly correlated, but, in this case, we can easily see that the correlation does not mean that ice cream causes crime or that crime inspires ice cream consumption.

The main reasons for correlations are:

  1. There can be a third variable that affects both of the observed variables. The outside temperature, in this case, might be the reason that increased crime and ice cream consumption seem related.
  2. There could be a confounding effect involved, such as the placebo effect in medical testing.
  3. The variables could be completely unrelated, or so tenuously related as to make the relationship meaningless, as in the case of the global warming and pirates.

So, you may wonder, why bring up power lines? In 1979, a study was published suggesting that there was a correlation between residence near “high-current configurations” of electrical wiring and death from cancer. You can imagine that this caused quite a stir, especially once a three part article was published in The New Yorker in 1989 implicating everything from power lines to electric blankets to video-display terminals as causing adverse health effects.

But the original study that started this all only showed a correlation, not a cause and effect, as people usually present it. While the original article written for the American Journal of Epidemiology does propose that the cancer rates were increased because of high current power lines, it concludes by suggesting other factors that could have influenced the results and notes that even though they show an increased risk of cancer, the risk is still very low. To paraphrase another blog, if your bank account jumps from $0.01 to $0.03, it’s a 200% increase, but it’s still not much money.

Many scientific studies show correlations, until the harder work of large, well-controlled studies can actually make conclusions about the causes. And, remember: any individual study could just be the odd man out and not accurately represent reality, which is why large replicable studies will always be important.

So to conclude, here are some of my favourite headlines that mistake correlation for causation:

And here’s a recent discussion of homicide rates from The Current. The second interviewee mentions correlations often, although his conclusions are arguable.

What are some of your favourites?

4 Responses to “Statistics for Skeptics Part 2 – Correlation vs. Causation”

  1. Fred says:

    “if your bank account jumps from $0.01 to $0.03, it’s a 300% increase, but it’s still not much money”

    Actually, that would be a 200% increase, but a good point none the less.

  2. Marlon says:

    Nice article and explanation of this tricky subject. I’m no mathematician but I think Fred may have led you astray. Took out my calculator and multiplied .01 x 300% and came up w/.03. Did the same for .01 x 200% and arrived at .02. I hope that doesn’t seem too nit-picky

  3. It’s all in how you word it, which is where I messed up the first time around.
    0.01 to 0.03 is an increase of 200% because you took 0.01 + 0.01 x 200%

    Another way to put it is that 0.03 is 0.01 x 300%, or 300% of 0.01.


  • Marion Kilgour

    Marion is a mechanical engineer, and also works to promote critical thinking and scientific literacy through local skeptical and atheist activism in Edmonton, Alberta. Marion especially wishes to encourage girls to consider science or technology-based careers, and is involved in the University of Alberta's Women in Scholarship, Engineering, Science and Technology (WISEST) project.