– David Morgan
Correlation and Causation
When presented with numerical data our brain, and secondary education, will seek out patterns and then, if required, plot and graph out the relationship. By the end of our school years we will do this with relative ease and confidently share conclusions that this data shows.
However just because a clear pattern emerges – have we satisfactorily established CAUSATION or simply identified a CORRELATION. Unfortunately the uncomfortable truth is that we never really understood the difference.
In his 2006 documentary, An Inconvenient Truth, Al Gore stood next to a huge graph that tracked the temperature of the earth and the concentration of carbon dioxide. His conclusion was that rising CO2 had caused a rise in temperature and the resulting climate changes. Over the next few years the counter-argument insisted that their was an undeniable correlation between the two variables but no causal link – indeed some observers argued the relationship was the inverse with temperature causing carbon dioxide rise. While this argument has since been resolved by the IPCC – the problem of causation and correlation remains.
Data in public policy is even more complex to analyse and the problems with rushing to judgement on the link between two data sets is commonplace. This is best exemplified by the predictions regarding inner city crime in the US during the nineties. Politicians who made predictions of a crime wave linked to poverty, unemployment and drug use were in abundance. However explaining the eventual drop in crime rates proved equally elusive. This is famously covered by Dubner and Levitt in Freakonomics who establish a clear relationship between abortion rates and crime.
The question for educators is whether we condition students to assume all relationships are causation without really considering correlation.
Many years ago I would have students investigate factors that would affect how long a candle would burn when covered by a glass beaker. Most would vary the size of the beaker and find that the bigger the beaker the larger volume of air would allow the candle to burn longer. Occasionally students would vary the size of candle and would struggle to come to a conclusion. Every once in a while a student would suggest the colour of the candle – as a caring teacher I would discourage them from this path. Was this the right thing to do? I now question that decision. I have no idea whether the coloured compounds in the wax has any effect on the rate of combustion. I know that there will be data and I know that there could even be a possible pattern in that data. Establishing whether that pattern is causation or correlation, however, would be a much greater test of a student’s scientific knowledge than the pre-determined and well known causation that the rest of the class has found.
“Just because two variables have a statistical relationship with each other does not mean that one is responsible for the other.” -Nate silver
No doubt we are teaching our students to collect and analyse data, but are we being too safe in our choice of experiments and investigations. By using the same old experiences we are conditioning them to expect and assume causation. Should’t we be having them look at data that is less obvious and have them argue over causation and correlation? There won’t always be a clearcut answer but isn’t that more like real life anyway?
David Morgan is an international teacher and administrator currently based in Switzerland
Gore, Al, Davis Guggenheim, Laurie David, Lawrence Bender, Scott Z. Burns, Jeff Skoll, Lesley d, Bob Richman, Jay Cassidy, Dan Swietlik, and Michael Brook. An Inconvenient Truth. Hollywood, Calif: Paramount, 2006.
Levitt, Steven D., and Stephen J. Dubner. Freakonomics. Harper Trophy, 2006.